It’s possible that I shall make an ass of myself. But in that case one can always get out of it with a little dialectic. I have, of course, so worded my proposition as to be right either way (K.Marx, Letter to F.Engels on the Indian Mutiny)
Russian neo-Nazi group 'Rusich' regularly posts images of war crimes
IDENTICAL IDEOLOGY AS UKR. NEO NAZI'S
Russian neo-Nazi paramilitary group Rusich took to Telegram in mid-November to post a photo of one of its fighters posing in front of three bodies of Ukrainian soldiers. It is thought to be the tip of the iceberg in terms of war crimes committed by this group, which is currently deployed on the Ukrainian front.
WARNING: SOME OF THE INFORMATION IN THIS ARTICLE MAY BE UPSETTING
In the photo, the Russian fighter, wearing a ski mask and a bulletproof vest and carrying a Kalashnikov, poses proudly in front of three bodies on the ground. The three dead men are all dressed in the military fatigues – complete with a yellow armband – worn by Ukrainian soldiers, though they have been stripped of their weapons. The photo, which was posted on Telegram on November 15, is captioned: “Take this as an example. That’s how an army of victors takes photos, not victims.”
According to pro-Russian blogger ButusovPlus, the photo was taken in the region of Pokrovsk, though it is impossible to independently verify this location.
A few hours later, the photo was followed by a second post, this time one that called on followers to send similar images in order to obtain an unspecified sum of cryptocurrency.
The message goes on to become even more macabre. “We are announcing a competition. The first three people who send a photo of prisoners who have clearly been executed will get a crypto money reward from Rusich.”
The post has since been deleted by the administrator of the Telegram channel.
The Telegram channel that has been sharing these photos is unfortunately well-known. Since it was created, it has featured content promoting war crimes, including decapitations, the humiliation of prisoners and executions.
These atrocities are the hallmark of Rusich, a group of Russian mercenaries who regularly display neo-Nazi references. They became well-known in Donbas in 2014, then acted alongside the Russian mercenary operation, the Wagner Group. Currently, members of this group are active on the Ukrainian front. While it is difficult to estimate how many of them there are, analysts agree that Rusich had a few dozen fighters in 2022. ‘Their aim is to terrorise their adversary’
Today, Rusich has become a byword for terror, says Candace Rondeaux, a professor at Arizona State University:
"Rusich is one of the pillars of the irregular warfare that the Kremlin has been engaged in since 2014. Their aim is to terrorise their adversaries by showing off their brutality.
Rusich was close to the Wagner universe and that of the Redut group (Editor’s note: a Russian mercenary group with close links to the intelligence services); they intervened in Syria but also in Libya. It is also a paramilitary group that operates with the blessing of the GRU (Editor’s note: the Russian military intelligence service) and recruits within the Russian ultra-nationalist movement and amongst former Russian parachutists (Editor’s note: also called the VDV).
Since it was created in 2014, Rusich has been seen as a force that carries out complex operations, including reconnaissance missions, sabotage or sniping. Some of their forces are operating behind enemy lines in Ukraine.” Neo-Nazi ideology
The members of the unit regularly show off ultra-nationalist and neo-Nazi symbols on Telegram. Case in point: the badge worn by members of the unit, which features an ultra-nationalist symbol called the Kolovrat superimposed on top of the Russian imperial flag (black, yellow and white).
Ricardo Parreira, an expert in far-right symbols, says that the Kolovrat has become a sort of banner for neo-Nazi groups in Eastern Europe:
“This is a symbol that was supposedly used by Slavic civilisations, as far back as the 11th century, but there are very few artefacts that document this. After the end of the Second World War, it was adopted by several neo-Nazi groups in Eastern Europe and in Russia, like the neo-Nazi party Russian National Unity.
When the war in Ukraine began in 2014, the Kolovrat symbol gained more recognition and was then picked up by international far-right groups. Both Russian and Ukrainian nationalist groups wore this symbol during the conflict, even though they were enemies."
The men in Rusich also wear other symbols favoured by neo-Nazis around the world. Rusich fighters often flaunt the Tyr and Odal runes from Norse mythology.
Parreira says that these symbols are used by many ultra-nationalist groups around the world:
"These days, all of these runes are references to a white supremacist and neo-Nazi ideology. But they were also used by Nazis, back in the day, as well as by some pagan movements, some of which are racist.
These symbols are meant to perpetuate the myth of a white people with Indo-European roots, supposedly the ancestors of both the Slavs and people from the Nordic countries. So they are symbols of an ethnopluralism (Editor’s note: a political ideology that promotes distinct ethnic and cultural groups living side by side but not mixing) that appears as frequently amongst the French extreme right (Institut Iliade, Nouvelle Droite) as in Russian supremacist circles."
Rusich members also seem to enjoy making references to the Third Reich. In a Telegram post from May 30, 2025, fighters from the unit wished one of their fellow fighters a happy birthday by adding an SS symbol to the wing of a drone.
One of the founders and the current leader of Rusich, Alexey Milchakov, stated publicly in an interview that he considers himself a Nazi and not a patriot or a nationalist.
He also called for the extermination of Ukrainians, “so they can no longer raise their children”. Ukrainian investigative group Molfar reported that Milchakov is believed to have cut off the ears of Ukrainian fighters in 2014 before publishing images of these atrocities on VKontakte.
According to the pro-Ukrainian ButusovPlus, Milchakov is actually deployed with the 417th Reconnaissance Battalion of the 42nd Guards Motor Rifle Division within the Russian army. According to the Institute for the Study of War, Milchakov’s unit was active to the west of the town of Zaporizhzhia in October 2025. War crimes posted on social media
Throughout its existence, the Rusich group’s cruelty has set it apart from other Russian mercenary groups. In 2017, one of its members posed proudly holding the severed head of a man near Palmyra, Syria.
The administrators of the Rusich Telegram channel continue to post scenes of horror. On November 26, 2024, the account posted an image of a severed head with a severed hand stuffed into its mouth. We don’t know where this image was taken.
In July 2024, the group posted an image of the body of a man lying in the middle of the street with a sword stuck into his chest. The body is partially dressed in Ukrainian Army military fatigues. An image that the account said was a “surprise for its followers”.
Similar photos and videos of crimes posted on social media were used in the trial of one of Rusich’s leaders, Yan Petrovsky. Petrovsky was arrested in Finland on July 20, 2023. He was sentenced to life in prison in Finland for four war crimes that he committed in eastern Ukraine in 2014.
‘Rusich perpetuates the idea that you can act in total impunity’
For Rondeaux, these posts are part of a well-planned communication strategy within the group:
“Communication has been part of the very DNA of Rusich since 2014. Back then, the group had its own Instagram account, its own YouTube channel and it was also very active on VKontakte (Editor’s note: a Russian social media network).
Rusich is carrying out psychological warfare on social media by publishing images of its crimes. It perpetuates the idea that the Kremlin and its troops can act in total impunity.
That serves a particular narrative, certainly now when Ukraine and Russia are considering a peace agreement that would not include any kind of system of reparations for Russian war crimes.”
Experts Agree: It Was Illegal to Follow Hegseth’s Illegal Orders
Under well-established law, those who complied with the orders cannot escape individual criminal responsibility for the killing of the two survivors in the event they are brought to trial. Secretary of Defense Pete Hegseth testifies during his Senate Armed Services confirmation hearing on Capitol Hill on January 14, 2025 in Washington, DC. (Photo by Andrew Harnik/Getty Images)
The question of when it is lawful for US military personnel to refuse an unlawful order has become a point of discussion in the political arena. Those conversations took a turn with the Washington Post and CNN reporting over Thanksgiving weekend that Secretary of Defense Pete Hegseth had issued a verbal order to “kill everyone” in the initial US military strike on suspected drug smugglers in the Caribbean, resulting in US special forces’ allegedly killing two shipwrecked survivors who were clinging to the wreckage of their vessel on Sept. 2, 2025.
In this article, we do not engage with the political discussion, but rather examine the law that applies to the alleged facts of the operation and Hegseth’s reported order. And with respect to the legal assessment of that operation, we will not be dealing with the broader question of whether the attack on the boat was unlawful as such, which it was (see articles published at Just Security by Marty Lederman; Michael Schmitt; and a podcast discussion with Tess Bridgeman, Brian Finucane, and Rebecca Ingber). Instead, we focus on a narrower aspect of the strike, the purported order to kill all aboard the vessel and the resulting second strike on the boat that killed the survivors.
As a matter of law, there are two central issues to address. The first concerns the circumstances in which military personnel have a duty to refuse to obey an order and, relatedly, whether a superior order can relieve them of criminal responsibility. The second is whether the orders in this case triggered that duty or provided those involved a defense. As both issues are context dependent, we begin with the facts.
The Reported Order(s) and Military Operation
Without rehashing the well-known and fairly straightforward reported chain of events on September 2, it is essential to understand that there were apparently two different orders in the military chain of command.
Secretary of Defense Pete Hegseth’s verbal order
The Washington Postreported: The longer the US surveillance aircraft followed the boat, the more confident intelligence analysts watching from command centers became that the 11 people on board were ferrying drugs.
Defense Secretary Pete Hegseth gave a verbal directive, according to two people with direct knowledge of the operation. “The order was to kill everybody,” one of them said.
Note that “Secretary of Defense Pete Hegseth had ordered the military prior to the operation to ensure the strike killed everyone on board, but it’s not clear if he knew there were survivors prior to the second strike, one of the sources said,” CNN reported.
The Washington Post reported: Two survivors were clinging to the smoldering wreck.
The Special Operations commander overseeing the Sept. 2 attack… ordered a second strike to comply with Hegseth’s instructions, two people familiar with the matter said. The two men were blown apart in the water.
Adm. Frank M. “Mitch” Bradley, told people on the secure conference call that the survivors were still legitimate targets because they could theoretically call other traffickers to retrieve them and their cargo, according to two people. He ordered the second strike to fulfill Hegseth’s directive that everyone must be killed.
This order implicates the duty of subordinate commanders and those executing the strike to refuse to comply with unlawful orders.
Following the strike, Hegseth told reporters, “We smoked a drug boat, and there’s 11 narco terrorists at the bottom of the ocean, and when other people try to do that, they’re going to meet the same fate.” Note that according to an earlier report by the New York Times, the targeted boat had “altered its course and appeared to have turned around before the attack started.”
The Duty to Refuse Unlawful Orders
From the perspective of those receiving them, unlawful orders raise two issues. The first is whether there is a duty to refuse them. The United States clearly imposes such a duty. In particular, the Department of Defense’s (DOD) Law of War Manual (2023) emphasizes the obligation, giving, as a paradigmatic example, an order to kill shipwrecked persons. 18.3.2.1 Clearly Illegal Orders to Commit Law of War Violations. The requirement to refuse to comply with orders to commit law of war violations applies to orders to perform conduct that is clearly illegal or orders that the subordinate knows, in fact, are illegal. For example, orders to fire upon the shipwrecked would be clearly illegal.
The Manual cautions, however, that “[s]ubordinates are not required to screen the orders of superiors for questionable points of legality, and may, absent specific knowledge to the contrary, presume that orders have been lawfully issued.” But in clear cases, the duty attaches. As the Manual for Courts-Martial explains, the general presumption that an order can be inferred to be lawful “does not apply to a patently illegal order, such as one that directs the commission of a crime.”
An even more granular explanation of the duty to refuse unlawful orders is provided in the US Navy-Marine Corps-Coast Guard Commanders Handbook on the Law of Naval Operations (§ 6.1.3.2): All naval personnel have a duty to comply with the law of armed conflict in good faith; prevent violations by others to the utmost of their ability; and refuse to comply with clearly illegal orders to commit violations of the law of armed conflict. Naval personnel have an affirmative obligation to promptly report violations which they become aware. When appropriate, naval personnel should ask questions through appropriate channels and consult with the command legal advisor on issues relating to the law of armed conflict. Naval personnel should adhere to regulations, procedures, and training, as these policies and doctrinal materials have been reviewed for consistency with the law of armed conflict. Commands and orders should not be understood as implicitly authorizing violations of the law of armed conflict where other interpretations are reasonably available.
These US duties track international law, for, as the International Committee of the Red Cross (ICRC) has asserted, under the law of armed conflict (LOAC), “[e]very combatant has a duty to disobey a manifestly unlawful order” (ICRC, Customary IHL study, Rule 154).
And refusal to obey an unlawful order is not an offense in the US armed forces. Under the Uniform Code of Military Justice, an offense occurs if the accused (1) violates or fails to obey any lawful general order or regulation; (2) having knowledge of any other lawful order issued by any member of the armed forces, which it is his duty to obey, fails to obey the order; or (3) is derelict in the performance of his duties (art. 92).
It is also an offense if a member of the armed forces “willfully disobeys a lawful command of his superior commissioned officer” (art. 90). Thus, the fact that an order is unlawful precludes conviction for its violation. So, although orders may generally be presumed lawful, if they are clearly unlawful, US military personnel have an affirmative duty to refuse them and may not be prosecuted for doing so.
No Defense of Superior Orders
The second issue raised by orders is whether they constitute a defense available to those acting unlawfully, but pursuant to them. It has long been the case under customary international law that “superior orders” is no defense for war crimes. The Charter of the International Military Tribunals at Nuremberg and Tokyo excluded the defense (arts. 8 and 6, respectively), as did the 1950 Nuremberg Principles (prin. IV). The absence of a superior orders defense has also been confirmed in the statutes of modern war crimes tribunals, including those of the International Criminal Court and the International Criminal Tribunals for the Former Yugoslavia and Rwanda (arts. 33, 7, and 6, respectively). Indeed, the defense is unavailable to international law violations generally. For instance, the United Nations Convention Against Torture and the Inter-American Convention on the Forced Disappearance of Persons prohibit superior orders as a defense in national legislation implementing their prohibitions (arts. 2 and VIII, respectively).
As with the affirmative duty to disobey an unlawful order, the ICRC has accurately stated that under customary international law, “[o]beying a superior order does not relieve a subordinate of criminal responsibility if the subordinate knew that the act ordered was unlawful or should have known because of the manifestly unlawful nature of the act ordered.” (ICRC Customary International Humanitarian Law study, Rule 155).
US military law likewise rejects the defense of superior order in the Manual for Courts-Martial. Rule 916(d) provides, “It is a defense to any offense that the accused was acting pursuant to orders unless the accused knew the orders to be unlawful or a person of ordinary sense and understanding would have known the orders to be unlawful.” The touchstone case reflecting the principle is US v. Calley, which dealt with the murder of 22 children, women, and old men in the South Vietnamese village of My Lai. Lt. Calley claimed he was obeying an order because “he had been taught the doctrine of obedience throughout his military career” and that he “was acting in ignorance of the laws of war.” The US Court of Military Appeals held that, the obedience of a soldier is not the obedience of an automaton. A soldier is a reasoning agent, obliged to respond, not as a machine, but as a person. The law takes these factors into account in assessing criminal responsibility for acts done in compliance with illegal orders.
The acts of a subordinate done in compliance with an unlawful order given him by his superior are excused and impose no criminal liability upon him unless the superior’s order is one which a man of ordinary sense and understanding would, under the circumstances, know to be unlawful, or if the order in question is actually known to the accused to be unlawful.
Thus, it is unlawful to obey an unlawful order, and merely following clearly illegal orders provides no defense. This being so, the questions in the September 2 strikes are whether Secretary Hegseth’s reported order to Adm. Bradley was clearly unlawful and whether Bradley’s apparent follow-on order to conduct the second strike was likewise manifestly unlawful.
What Law Applied to the Reported Orders?
Much attention has been focused on the laws of war as they may relate to the Hegseth order and resulting operation. In that regard, we must emphasize that LOAC did not apply to the September 2 strikes, because, as has been explained in multiple Just Security articles referenced above, the United States is not in an armed conflict with any drug trafficking cartel or criminal gang anywhere in the Western Hemisphere. There is no international armed conflict because, inter alia, there are neither hostilities between States nor the requisite degree of State control over alleged drug cartels operating the boats. And there is no non-international armed conflict, both because the cartels concerned do not qualify as organized armed groups in the LOAC sense, and because there were no hostilities between the United States and the cartels on September 2, let alone hostilities that would reach the requisite level of intensity to cross the armed conflict threshold. For the same reason, the individuals involved have not committed war crimes.
However, the duty to refuse clearly unlawful orders—such as an order to commit a crime—is not limited to armed conflict situations to which LOAC applies. Nor is rejection of a defense of superior orders restricted to war crimes. In fact, the more restrictive rules of international human rights law applied instead. As will be explained, the alleged Hegseth order and special forces’ lethal operation amounted to unlawful “extrajudicial killing” under human rights law (see also here). The federal murder statute would also apply, whether or not there is an armed conflict. (See, e.g., Marty Lederman’s analysis).
That said, the administration has reported to Congress, stated publicly, and recorded in legal and operational memoranda that it believes one or multiple “non-international armed conflicts” exist between the United States and 24 organizations in Latin America (whether it views the situation as one armed conflict, 24 separate ones, or some other combination is unclear). This being so, before turning to the law that was actually violated through the September 2 and subsequent operations, allow us to counterfactually consider the law that would apply had the administration been correct in characterizing the operation as occurring during an armed conflict.
The Prohibition of Ordering Denial of Quarter or Denying Quarter
Assuming solely for the sake of discussion that there was a non-international armed conflict at the time of the September 2 strikes, the most relevant LOAC rule applicable to the Hegseth and Bradley orders is the “denial of quarter,” i.e., an instruction not to allow any survivors (see, e.g., Working Group of Former Judge Advocates Generals’ statement on the Hegseth order).
The status of the prohibition on the denial of quarter (and on ordering or threatening its denial) was settled well over a century ago. It is applicable in both international and non-international armed conflicts as a matter of customary international law (ICRC, Customary International Humanitarian Law study, Rule 46). This is so with respect to its status as a violation of LOAC entailing the responsibility of the State concerned and as a war crime by the individuals issuing orders to deny quarter or carrying them out. We need not repeat here the major international texts and tribunal decisions that support that conclusion. One of us (Schmitt) walked through all of the relevant texts, from the US Civil War’s Lieber Code to the present, in a 2023 essay concerning a “kill everyone” order by the head of Russia’s Wagner Group (coauthored with LtCol John Tramazzo).
Here, suffice it to note that the DOD Law of War Manual is categorical: “It is… prohibited to conduct hostilities on the basis that there shall be no survivors, or to threaten the adversary with the denial of quarter. This rule is based on both humanitarian and military considerations.” The Manual further emphasizes that the rule “also applies during non-international armed conflict” (§ 5.4.7).
A closely related prohibition implicated in the September 2 strikes, which also applies in both international and non-international armed conflict, is on attacking those who are hors de combat, a condition that includes those who are “defenseless” because they are shipwrecked (see ICRC Customary International Humanitarian Law study, rule 47 and related practice). As the DOD Law of War Manual explains (§ 5.9.4), Shipwrecked combatants include those who have been shipwrecked from any cause… Persons who have been incapacitated by… shipwreck are in a helpless state, and it would be dishonorable and inhumane to make them the object of attack. In order to receive protection as hors de combat, the person must be wholly disabled from fighting.
The Commander’s Handbook on the Law of Naval Operations similarly provides, “Intentional attack on a combatant who is known to be hors de combat constitutes a grave breach of the law of armed conflict” (§ 8.2.3). Indeed, as noted in the Newport Manual on the Law of Naval Warfare published by the US Naval War College’s Stockton Center, Geneva Convention II sets forth a legal framework for the humane treatment and protection of victims of armed conflict at sea. The Convention requires parties to the conflict to, inter alia, respect and protect individuals falling within the scope of the Convention “who are at sea and who are wounded, sick or shipwrecked.” Parties to a conflict are thus required, after each engagement and without delay, to “take all possible measures to search for and collect the shipwrecked, wounded and sick,” without discriminating between their own and enemy personnel.
To be clear, there is no exception to the prohibition on attacking those who are hors de combat due to being shipwrecked because they might escape or otherwise receive rescue assistance from their forces. The only basis for treating them as subject to continued attack is if they are, in fact, not hors de combat because they continue to fight. Doctrine and Prosecutions on Denial of Quarter
This analysis of the LOAC rules merits being supplemented with three additional points. First, each US servicemember has an obligation to report evidence that any US operation potentially involved killing shipwrecked survivors or a denial of quarter. According to the Commander’s Handbook on the Law of Naval Operations (§ 6.3; see also DOD Directive 2311.01): All military and US civilian employees, contractor personnel, and subcontractors assigned to or accompanying a DOD component must report through their chain of command all reportable incidents, including those involving allegations of non-DOD personnel having violated the law of war.
Examples of incidents that “must be reported” include: (1) “Offenses against the Wounded, the Sick, [and] Survivors of Sunken Ships,” such as “willfully killing”; (2) “Other Offenses against Survivors of Sunken Ships,” including, “when military interests permit, failure to search out, collect, make provision for the safety of, or to care for survivors;” and (3) “Denial of quarter, unless bad faith is reasonably suspected” (§ 6.3).
Second, a landmark 1921 case emerging out of World War I clearly set forth the rule that killing shipwrecked survivors of a boat strike is a war crime and that superior orders offer no defense to such conduct, because such orders must be disobeyed. In the Llandovery Castle case, the Imperial Court of Justice considered a June 1918 incident after a German U-boat sank the Llandovery Castle, a Canadian hospital ship. The U-boat Commander claimed he thought the ship was carrying American airmen. In convicting the defendants for firing on the survivors who were in lifeboats, the court noted that by that point, the international legal prohibition on killing survivors of a maritime attack was manifest. The firing on the boats was an offence against the law of nations. In war on land the killing of unarmed enemies is not allowed (compare the Hague regulations as to war on land, para. 23(c)), similarly in war at sea, the killing of shipwrecked people, who have taken refuge in life-boats, is forbidden. … The fact that his deed is a violation of international law must be well-known to the doer, apart from acts of carelessness, in which careless ignorance is a sufficient excuse. In examining the question of the existence of this knowledge, the ambiguity of many of the rules of international law, as well as the actual circumstances of the case, must be borne in mind, because in war time decisions of great importance have frequently to be made on very insufficient material. This consideration, however, cannot be applied to the case at present before the court. The rule of international law, which is here involved, is simple and is universally known. No possible doubt can exist with regard to the question of its applicability. (emphasis added)
Accordingly, the court held that the German crew could not claim to be following orders as a defense because such an order would be clearly unlawful: It is certainly to be urged in favor of the military subordinates, that they are under no obligation to question the order of their superior officer, and they can count upon its legality. But no such confidence can be held to exist, if such an order is universally known to everybody, including also the accused, to be without any doubt whatever against the law. This happens only in rare and exceptional cases. But this case was precisely one of them, for in the present instance, it was perfectly clear to the accused that killing defenceless people in the life-boats could be nothing else but a breach of the law. … They should, therefore, have refused to obey. As they did not do so, they must be punished.“ (emphasis added)
The DOD Law of War Manual cites and quotes the Llandovery Castle case to illustrate the point that clearly illegal orders must be refused (see DOD Law of War Manual, § 18.3.2.1).
Notably, in its sentencing assessment, the court stated that “the principal guilt rests with” the U-boat Commander who issued the order, while his subordinates could obtain some mitigation of sentence given the pressure entailed in refusing a military order.
Third, a famous World War II case involved a similar set of facts. In the 1945 Peleus Trial, a British Military Court sitting in Hamburg considered a March 1944 incident in which a German submarine sank a Greek ship chartered by the British Ministry of War Transport. Upon the orders of the German commander Heinz Eck, the U-boat members fired a machine gun and threw grenades at Peleus’ crew members who had survived the first attack but were shipwrecked in the water. The prosecutor and the judge advocate (who at that time served as the court’s legal adviser) both relied on the Llandovery Castle case. In response to the defendants’ plea of superior orders, the judge advocate stated: The duty to obey is limited to the observance of orders which are lawful. There can be no duty to obey that which is not a lawful order. …
It is quite obvious that no sailor and no soldier can carry with him a library of international law, or have immediate access to a professor in that subject who can tell him whether or not a particular command is a lawful one. If this were a case which involved the careful consideration of questions of international law as to whether or not the command to fire at helpless survivors struggling in the water was lawful, you might well think it would not be fair to hold any of the subordinate accused in this case responsible for what they are alleged to have done; but is it not fairly obvious to you that if in fact the carrying out of Eck’s command involved the killing of these helpless survivors, it was not a lawful command, and that it must have been obvious to the most rudimentary intelligence that it was not a lawful command, and that those who did that shooting are not to be excused for doing it upon the ground of superior orders? (emphasis added)
The court sentenced Eck and two other defendants to death, another to life imprisonment, and the fifth defendant to 15 years imprisonment.
Assuming the facts as reported about the September 2 strike, and if LOAC and war crimes law had applied (they do not), Secretary Hegseth and Admiral Bradley’s orders were self-evidently unlawful because they ordered no quarter. Moreover, the second strike on the boat would qualify as an attack on those shipwrecked persons who are hors de combat. Whether Secretary Hegseth knew there were shipwrecked survivors is unclear, but Admiral Bradley reportedly did and ordered their attack anyway.
If those involved believed they were engaged in an armed conflict, we find it difficult to imagine they could not have known that the orders were unlawful. The more military training and experience they have, the more implausible such a claim is. Applying International Human Rights Law to the Alleged Facts
The law of armed conflict is generally a more permissive legal regime for the use of military force than international human rights law (IHRL). In particular, the LOAC permits targeting members of the armed forces, including members of organized armed groups, based on their status, and others if and for such time as they “directly participate in hostilities,” which encompasses more than conducting attacks. By contrast, targeting based on status outside an armed conflict is prohibited. Acts opening the door to the use of force against an individual are generally limited to situations in which they pose an imminent threat of death or grievous bodily harm. If the Hegseth and Bradley orders and the ensuing second strike had been violations of LOAC in a non-international armed conflict, they would, a fortiori, have violated human rights law as a matter of peacetime law enforcement.
With respect to the US lethal strikes on suspected drug trafficking vessels at issue here, two of us (Schmitt and Goodman, along with coauthor Marko Milanovic) have explained why “there is absolutely no question that the US lethal strikes on the boats are a violation of international human rights law.” Without rehashing that analysis here, the bottom line is that the US strikes on suspected drug traffickers at sea are clearly arbitrary deprivations of the right to life under IHRL, an obligation that the United States acknowledges applies extraterritorially. As they wrote: The widely-accepted standard for arbitrariness prohibits the use of force likely to cause death or grievous bodily injury “except in self-defence or defence of others against the imminent threat of death or serious injury, to prevent the perpetration of a particularly serious crime involving grave threat to life, to arrest a person presenting such a danger and resisting their authority, or to prevent his or her escape, and only when less extreme means are insufficient to achieve these objectives” (Basic Principles on the Use of Force and Firearms by Law Enforcement Officials; see also UN Human Rights Committee, General Comment 36, para 12).
If the lack of an imminent threat of death or serious injury on the part of individuals suspected of trafficking drugs at sea (quite plausibly here, ferrying cocaine from Venezuela to a transhipment point for onward distribution in Europe) is obvious with respect to the campaign as a whole, it is doubly so with respect to a vessel that, as has been reported, had turned around prior to the US strike. It is even more patently obvious that it is an arbitrary deprivation of the right to life—i.e., murder—to fire on the shipwrecked survivors of that strike, as has now been reported.
In sum, there is simply no plausible argument that the reported killing of two survivors clinging to the burning wreckage of their stricken vessel could be anything other than an extrajudicial killing. It is equally clear that, according to long-standing law (including prevailing US legal interpretations), the reported Hegseth and Bradley orders to fire on them were manifestly unlawful, and that those carrying out that order cannot rely on a superior orders defense if prosecuted for those actions due to the egregious illegality of the order.
Concluding Thoughts
The September 2 strikes on the purported drug boat neither violated the law of armed conflict nor amounted to war crimes, because they did not occur during an armed conflict. However, if the facts are as reported, there is little question that the order by Secretary Hegseth and the ensuing order by Admiral Bradley to conduct the second strike were unlawful, because the killing of the two survivors was a serious violation of international human rights law.
Moreover, both orders were clearly unlawful. Under well-established law, those who complied with the orders cannot escape individual criminal responsibility for the killing of the two survivors in the event they are brought to trial in a US military court-martial, a federal trial, or a domestic criminal proceeding in another State that has jurisdiction, for instance, based on the nationality of the victims. If actually issued, these orders irresponsibly and unlawfully placed all those involved in the attack in serious legal jeopardy. If the reporting is accurate, those orders should, as a matter of law, have been refused. Editor’s note: Readers may also be interested in Jeremy Chin, Margaret Lin, and Aidan Arasasingham, Timeline of Vessel Strikes and Related Actions
Michael Schmitt Michael Schmitt is professor of international law at the University of Reading, affiliate at Harvard Law School’s Program on International Law and Armed Conflict, and visiting research professor at the International Institute of Humanitarian Law. He formerly served as the G. Norman Lieber distinguished scholar at West Point, chair of the Stockton Center for International Law at the US Naval War College, where he is pofessor emeritus, dean of the George C. Marshall Center European Center for Security Studies, and professor of law at the University of Exeter, Durham University, and the United States Air Force Academy. Professor Schmitt is a retired US Air Force judge advocate, having specialized in international and operational law. He is the general editor of the Lieber Studies series (OUP) and sits on many international law advisory and editorial boards. Full Bio >
Ryan Goodman Ryan Goodman is the Anne and Joel Ehrenkranz professor of law at New York University school of law, specialising in international human rights. Follow him on Twitter @rgoodlaw Full Bio >
Tess Bridgeman Dr. Tess Bridgeman is coeditor-in-chief of Just Security. She served in the White House as special assistant to the president, associate counsel to the president, and deputy legal adviser to the National Security Council (NSC) during the Obama administration. She also served at the State Department in the Office of the Legal Adviser as special assistant to the legal adviser and, prior to that role, as an attorney adviser in the Office of Political-Military Affairs. Full Bio >
Monday, December 01, 2025
The mystery of the missing deep ocean carbon fixers
UCSB study reshapes understanding of deep-ocean carbon storage with implications for long-term climate stability
(Santa Barbara, Calif.) — In a step toward better understanding how the ocean sequesters carbon, new findings from UC Santa Barbara researchers and collaborators challenge the current view of how carbon dioxide is “fixed” in the sunless ocean depths. UCSB microbial oceanographer Alyson Santoro and colleagues, publishing in the journal Nature Geoscience, present results that help to reconcile discrepancies in accounting for nitrogen supply and dissolved inorganic carbon (DIC) fixation at depth.
“Something that we’ve been trying to get a better handle on is how much of the carbon in the ocean is getting fixed,” Santoro said. “The numbers work out now, which is great.”
This project was supported in part by the National Science Foundation.
Who’s doing the fixing? The ocean is the Earth’s largest carbon sink, buffering us from the worst that climate change can throw at us by absorbing a third of our carbon dioxide emissions, which in turn regulates global temperatures. We rely heavily on the ocean for this phenomenon, which is why it’s important to fully understand the complex processes that enable it.
“We want to know how carbon moves around the deep ocean, because in order for the ocean to impact the climate, carbon has to make it from the atmosphere to the deep ocean,” Santoro said.
In the ocean, the majority of this inorganic carbon fixing work is conducted by microbes. Phytoplankton, a group of single-celled organisms that take up inorganic carbon dioxide (including dissolved carbon dioxide gas) at the ocean’s surface, are known as autotrophs. They produce their own food in the same way plants on land photosynthesize carbon dioxide and water, producing organic matter (sugars) and oxygen.
The prevailing idea has been that while most DIC fixation happens in the upper, sunlit layer thanks to photosynthetic phytoplankton, a significant amount of non-photosynthetic DIC fixation also occurs in the “dark” layers of the ocean, an assimilation dominated mainly by autotrophic archaea that evolved to oxidize ammonia (a nitrogen-containing compound) for energy rather than sunlight.
However, when tracking these carbon fixing microbes’ nitrogen energy budget through water column sampling, researchers soon found that the numbers weren’t matching up.
“There was a discrepancy between what people would measure when they went out on a ship to measure carbon fixation and what was understood to be the energy sources for microbes,” Santoro said. “We basically couldn’t get the budget to work out for the organisms that are fixing carbon.” They needed energy to do that, she explained, but there didn’t seem to be enough available nitrogen-based energy to go around in the deep ocean for the rates of carbon fixation being reported throughout the water column.
This mystery has long been on the minds of Santoro and the paper’s lead author Barbara Bayer, who have been working to fill this gap in our understanding of the ocean’s carbon cycle for almost a decade. Previous work has explored the hypothesis that maybe these carbon-fixing archaea were more efficient at their jobs than assumed, requiring less nitrogen to fix carbon, though their results indicated that was not the case.
For this paper, the team took a different approach, asking instead how big the contribution of these ammonia oxidizers was to the total dissolved inorganic carbon fixation rates in the dark ocean. To find out, Bayer devised a clever experiment.
“She came up with a way to specifically inhibit their activity in the deep ocean,” Santoro explained. By restricting the oxidizers with a special chemical, she continued, the rate of carbon fixing should be drastically reduced. The inhibitor, phenylacetylene, was confirmed to have no other measurable effects on other community processes.
Their results indicated that despite inhibiting these ammonia oxidizers — mostly archaea that are abundant in the dark ocean — the rate of carbon fixation in the study areas didn’t drop as much as expected.
So if not the ammonia-oxidizing archaea, then who could be doing the carbon fixing in the depths? The list of suspects has grown to include other microbes in the neighborhood, particularly bacteria and some archaea.
“We think that what this means is that the heterotrophs — microorganisms that feed on organic carbon from decomposing microbes and other marine life — are taking up a lot of inorganic carbon in addition to the organic carbon that they usually consume,” Santoro said, “meaning that they’re also responsible for fixing some carbon dioxide.
“And that’s really interesting because even though we know this to be a theoretical possibility, we didn’t really have a quantitative number on what fraction of the carbon in the deep ocean was getting fixed by these heterotrophs versus autotrophs. And now we do.”
These findings also help to paint a clearer picture of how the deep ocean’s food web works.
“There are basic aspects of how the food web works in the deep ocean that we don’t understand,” Santoro said, “and I think of this as figuring out how the very base of the food web in the deep ocean works.”
More mysteries of the deep
Further work in this realm for Santoro and her collaborators will dive into the finer aspects of carbon fixation in the ocean, such as how the nitrogen cycle and carbon cycle interact with other elemental cycles in the ocean, including for iron and copper.
“The other thing we’re trying to figure out is once these organisms fix the carbon into their cells, how does it become available to the rest of the food web?” she noted. “What kinds of organic compounds might they be leaking out of their cells that could be feeding the rest of the food web with?”
Research in this paper was also conducted by Nicola L. Paul, Justine B. Albers and Craig A. Carlson at UCSB; Katharina Kitzinger and Michael Wagner at the University of Vienna as well as Mak A. Saito at Woods Hole Oceanographic Institution.
Journal
Nature Geoscience
Article Title
Minor contribution of ammonia oxidizers to inorganic carbon fixation in the ocean
Small changes make some AI systems more brain-like than others
For visual AI systems, selecting the right blueprint may accelerate learning
Artificial intelligence systems that are designed with a biologically inspired architecture can simulate human brain activity before ever being trained on any data, according to new research from Johns Hopkins University.
The findings, published in Nature Machine Intelligence, challenge conventional approaches to building AI by prioritizing architectural design over the type of deep learning and training that takes months, costs billions of dollars and requires thousands of megawatts of energy.
“The way that the AI field is moving right now is to throw a bunch of data at the models and build compute resources the size of small cities. That requires spending hundreds of billions of dollars. Meanwhile, humans learn to see using very little data,” said lead author Mick Bonner, assistant professor of cognitive science at Johns Hopkins University. “Evolution may have converged on this design for a good reason. Our work suggests that architectural designs that are more brain-like put the AI systems in a very advantageous starting point.”
Bonner and a team of scientists focused on three classes of network designs that AI developers commonly use as blueprints for building their AI systems: transformers, fully connected networks, and convolutional networks.
The scientists repeatedly modified the three blueprints, or the AI architectures, to build dozens of unique artificial neural networks. Then, they exposed these new and untrained AI networks to images of objects, people, and animals and compared the models’ responses to the brain activity of humans and primates exposed to the same images.
When transformers and fully connected networks were modified by giving them many more artificial neurons, they showed little change. Tweaking the architectures of convolutional neural networks in a similar way, however, allowed the researchers to generateactivity patterns in the AI that better simulated patterns in the human brain.
The untrained convolutional neural networks rivaled conventional AI systems, which generally are exposed to millions or billions of images during training, the researchers said, suggesting that the architecture plays a more important role than researchers previously realized.
“If training on massive data is really the crucial factor, then there should be no way of getting to brain-like AI systems through architectural modifications alone,” Bonner said. “This means that by starting with the right blueprint, and perhaps incorporating other insights from biology, we may be able to dramatically accelerate learning in AI systems.”
Next, the researchers are working on developing simple learning algorithms modeled after biology that could inform a new deep learning framework.
Convolutional architectures are cortex-aligned de novo
Asia PGI and partners unveil preview of PathGen: New AI-powered outbreak intelligence tool
Asia-led, “sovereign-by-design” platform built for secure, decentralised pathogen intelligence-sharing across borders aims to break data silos and provide faster “time to actionable insight” of outbreaks, from detection to control measures
From left to right. Seated: Mr Ng Boon Heong, Executive Director & Chief Executive Officer, Temasek Foundation; Ms Ho Ching, Chairman, Temasek Trust; Mr Ong Ye Kung, Minister for Health and Coordinating Minister for Social Policies; Ms Jennie Chua, Chairman, Temasek Foundation; Mr Goh Yew Lin, Chair, Governing Board, Duke-NUS Medical School.
Standing: Dr Lee Fook Kay, Head, Pandemic Preparedness, Temasek Foundation; Prof Vernon Lee, Chief Executive, Communicable Diseases Agency; Ms Zeng Xiaofan, Senior Program Officer, Gates Foundation; Prof Patrick Tan, Dean-designate, Duke-NUS Medical School; Prof Thomas Coffman, Dean, Duke-NUS Medical School; Prof Paul Pronyk, Director, Duke-NUS Centre for Outbreak Preparedness at the preview of PathGen // Image credit: Duke-NUS Medical School
SINGAPORE, 1 December 2025 – Asia Pathogen Genomics Initiative (Asia PGI) today offered the first public preview of PathGen, an AI-powered sense-making and decision-making support platform of pathogen genomics and contextual data. Designed for public health practitioners, clinicians and industry, it can help detect emerging disease threats earlier, assess risks faster, and coordinate responses within and across borders, all without compromising countries’ ownership of their respective sovereign data. The objective is to strengthen health security across Asia and beyond, reducing lives lost and livelihoods disrupted, and the economic impacts of communicable diseases.
The preview demonstration, hosted by the Duke-NUS Centre for Outbreak Preparedness (COP) and Temasek Foundation, showcased how PathGen could integrate diverse data sources – pathogen genomics, clinical information, population data, climate, and mosquito habitat patterns – powered by the latest AI technology and foundation models to provide enhanced situational awareness and decision-making support through timely, high-quality, actionable insights. The result: Faster decisions taken with higher resolution and greater confidence – for example guiding decisions on when to adjust treatment protocols, where to deploy vaccines, and how to allocate resources, before outbreaks spiral out of control.
More than 100 attendees, comprising senior health officials from the region, as well as philanthropic, scientific and technology partners, were at the preview and discussed governance, strategy, and next steps for regional deployment. In a symbolic show of regional commitment, partners from Indonesia, Malaysia, Singapore, Thailand, and Vietnam placed their hands on the PathGen logo to affirm their pledge to co-create PathGen as a shared public good for regional health security, while partners from the Philippines joined virtually. Mr Ong Ye Kung, Singapore’s Minister for Health and Coordinating Minister for Social Policies was the guest-of-honour at the PathGen preview.
PathGen is housed by Asia PGI which is led by the Duke-NUS Centre for Outbreak Preparedness. A coordination and capacity development hub advancing pathogen genomics sequencing for early detection, control and elimination of infectious diseases in the region, Asia PGI convenes more than 50 government and academic partners across 15 countries, with Singapore as its nerve centre. Asia PGI and PathGen are propelled by three key catalytic funders – the Gates Foundation, Temasek Foundation, and Philanthropy Asia Alliance. Four development partners – Amazon Web Services (AWS), IXO, Sequentia Biotech, and Sydney ID at the University of Sydney – are contributing core technologies and expertise to bring PathGen from concept to practice.
Why now?
Traditional epidemiological surveillance reports what is happening (e.g., counts of disease cases, hospitalisations, etc), while genomic surveillance has the potential to reveal the “who, where, and how” infections emerge in populations. But today’s systems are often hindered by silo-ed, non-interoperable databases that do not share information with one another, limited contextual data, data sovereignty barriers and policy constraints that slow responses and limit preparedness. Urgent action is needed to overcome these barriers as rapid population growth, unprecedented mobility, climate disruptions and antimicrobial resistance are driving more frequent and complex disease outbreaks across human and animal populations.
Recent AI breakthroughs which underpin PathGen allow better synthesis of genomic, clinical, population, environmental, and mobility data to help timely clinical and public health decisions. A “federated”, “sovereign-by-design” platform such as PathGen shares only the analytics; the underlying raw data is not moved or centralised to one location and remains under the control of the respective country/owner, enabling cooperation without compromising data integrity or eroding trust.
Professor Paul Pronyk, Director of Duke-NUS’ Centre for Outbreak Preparedness said: “This proof of concept shows how AI and pathogen genomics can work together to provide actionable intelligence for clinicians and public health authorities. By sharing only essential insights, countries can respond faster to outbreaks while strengthening trust and sovereignty.”
Dr Lee Fook Kay, Head of Pandemic Preparedness, Temasek Foundation said, “Every delay between detecting a pathogen and making the right public health decision costs lives. Temasek Foundation is catalysing PathGen, as it can integrate genomic information with other relevant surveillance, population and environmental data sources into timely insights that health authorities can act upon. A shared intelligence system that protects sovereignty, cuts response time, and stops outbreaks before they become crises – that’s the future of health security and preparedness!”
Shaun Seow, CEO of Philanthropy Asia Alliance, added: “PathGen fills a critical gap with a decision-support platform built for Asia’s needs and complexities. It enables shared intelligence without compromising data sovereignty, helping us better prepare for the next pandemic. Through our Health for Human Potential Community, we’re proud to support this effort to strengthen public health resilience across the region.”
What’s next?
PathGen will advance from proof-of-concept towards a launch-ready platform over the next 18 months, with pilots from early 2026 and a staged roll-out through 2027.
These efforts will be supported through the Asia PGI network of country partners. Country-level engagement aims to help define priorities and technical needs; establish plans for secure in-country deployment; set governance and benefit-sharing arrangements; deliver core analytics and decision support tools with integration to national systems; build capacity for public health laboratories and implementation teams; and provide regular briefings and demonstrations to align partners on strategy, governance, and next steps.
Information on PathGen’s development will be updated on the PathGen website.
###
About Duke-NUS Medical School
Duke-NUS is Singapore’s flagship graduate entry medical school, established in 2005 with a strategic, government-led partnership between two world-class institutions: Duke University School of Medicine and the National University of Singapore (NUS). Through an innovative curriculum, students at Duke-NUS are nurtured to become multi-faceted ‘Clinicians Plus’ poised to steer the healthcare and biomedical ecosystem in Singapore and beyond. A leader in ground-breaking research and translational innovation, Duke-NUS has gained international renown through its five signature research programmes and 10 centres. The enduring impact of its discoveries is amplified by its successful Academic Medicine partnership with Singapore Health Services (SingHealth), Singapore’s largest healthcare group. This strategic alliance has led to the creation of 15 Academic Clinical Programmes, which harness multi-disciplinary research and education to transform medicine and improve lives.
Temasek Foundation supports a diverse range of programmes that uplift lives and communities in Singapore and Asia. Temasek Foundation’s programmes are made possible through philanthropic endowments gifted by Temasek, as well as gifts and other contributions from other donors. These programmes strive to deliver positive outcomes for individuals and communities now, and for generations to come. Collectively, Temasek Foundation’s programmes strengthen social resilience; foster international exchange and catalyse regional capabilities; advance science; and protect the planet.
[From left to right] Mr Maximilian Fiedler, Regional CEO Asia Pacific and Managing Director Schaeffler (Singapore); Mr Uwe Wagner, CTO, Schaeffler AG; Dr Tan See Leng, Minister for Manpower and Minister-in-charge of Energy and Science & Technology, Ministry of Trade and Industry; Prof Lam Khin Yong, Vice President (Industry), NTU and Prof Christian Wolfrum, Deputy President and Provost, NTU
Nanyang Technological University, Singapore (NTU Singapore) and the leading global Motion Technology company Schaeffler have officially launched the next phase of their corporate laboratory partnership to drive research and innovation in AI-enabled humanoid robotics.
Gracing the launch of the new Schaeffler-NTU Corporate Lab: Intelligent Mechatronics Hub today as Guest of Honour was Dr Tan See Leng, Minister for Manpower and Minister-in-charge of Energy and Science & Technology, Ministry of Trade and Industry.
Located on NTU Singapore’s campus, the new 900-square-metre facility will contribute to Singapore’s strategic goal of strengthening advanced manufacturing and robotics. It marks another milestone in the collaboration between NTU and Schaeffler, which started in 2017.
The corporate laboratory is supported by the National Research Foundation, Singapore (NRF) under the Research, Innovation and Enterprise (RIE) 2025 plan, and developed in partnership with the Singapore Economic Development Board (EDB).
It will focus on advancing collaborative robotics, autonomous mobile robot platforms and assistive robotic systems, targeting applications in manufacturing, logistics and healthcare.
The lab will also collaborate with researchers from other institutions of higher learning, further reinforcing Singapore’s position as a regional hub for intelligent automation and humanoid robotics innovation.
It is part of Schaeffler’s global Schaeffler Hub for Advanced Research (SHARE) network that collaborates with leading universities worldwide through its company-on-campus concept.
Uwe Wagner, Chief Technology Officer at Schaeffler AG, said: “The next phase of collaboration at the Schaeffler Hub for Advanced Research at NTU marks a significant milestone in our long-standing partnership and reinforces our commitment to pioneering innovation in robotics and artificial intelligence. With a focus on advancing technologies for humanoid robotics, this partnership represents a key step forward in our holistic agenda to drive progress in this future field. Drawing on expertise across our eight product families, Schaeffler is best equipped to shape the future of humanoid robotics. By working closely with leading researchers at NTU, we strive to accelerate development and deliver value that resonates far beyond the regional level.”
ProfLam Khin Yong, NTU Vice President (Industry), added: “This expanded collaboration with Schaeffler reinforces NTU's position as a leading research university with strong multi-party partnerships between academia, industry, and public agencies. The corporate lab provides a platform for our researchers, doctoral candidates, and students to work on challenges in robotics alongside industry experts. We have also collaborated closely with Schaeffler engineers to develop robots that can co-work with humans, with advanced sensors improving sensitivity and safety, which has direct industrial impact. I’m confident that our innovations can boost the manufacturing sector and shape the future of autonomous and assistive robotics in Singapore and beyond.”
Cindy Koh, Executive Vice President, EDB said: “Schaeffler's continued investments in Singapore have contributed important capabilities to our advanced manufacturing ecosystem, and created highly skilled research, engineering and corporate jobs. The expanded corporate lab builds on the success of Schaeffler's longstanding partnership with Singapore's research community and universities, helping to connect academic research with real-world industry applications. This aligns with Singapore’s strategic interest to increase adoption of robotics and embodied AI in advanced manufacturing and unlock new opportunities across industries.”
Since the collaboration began in 2017, the NTU–Schaeffler partnership has produced numerous innovations.
These include a real-time visualisation of touch and force technology that enhances the precision and safety of robots in industrial settings through real-time sensing, and a universal soft gripper that can handle a wide range of objects with diverse geometries, stiffness levels, and surface properties to boost productivity and efficiency in manufacturing and supply chain applications.
Beyond research, the partnership supports talent development by training PhD, Master’s, and undergraduate students, providing them with hands-on experience through working alongside Schaeffler engineers and researchers on real-world projects. Many alumni of the programme have since assumed leadership positions in academia and industry, contributing to Singapore’s deep technology ecosystem and advanced manufacturing sectors.
SHARE at NTU will further enhance Schaeffler’s innovation footprint in Asia and support NTU’s continued drive for interdisciplinary research and industry collaboration to address some of the world’s most critical challenges.
How does AI think? KAIST achieves first visualization of the internal structure behind AI decision-making
The Korea Advanced Institute of Science and Technology (KAIST)
Although deep learning–based image recognition technology is rapidly advancing, it still remains difficult to clearly explain the criteria AI uses internally to observe and judge images. In particular, technologies that analyze how large-scale models combine various concepts (e.g., cat ears, car wheels) to reach a conclusion have long been recognized as a major unsolved challenge.
KAIST (President Kwang Hyung Lee) announced on the 26th of November that Professor Jaesik Choi’s research team at the Kim Jaechul Graduate School of AI has developed a new explainable AI (XAI) technology that visualizes the concept-formation process inside a model at the level of circuits, enabling humans to understand the basis on which AI makes decisions.
The study is evaluated as a significant step forward that allows researchers to structurally examine “how AI thinks.”
Inside deep learning models, there exist basic computational units called neurons, which function similarly to those in the human brain. Neurons detect small features within an image—such as the shape of an ear, a specific color, or an outline—and compute a value (signal) that is transmitted to the next layer.
In contrast, a circuit refers to a structure in which multiple neurons are connected to jointly recognize a single meaning (concept). For example, to recognize the concept of cat ear, neurons detecting outline shapes, neurons detecting triangular forms, and neurons detecting fur-color patterns must activate in sequence, forming a functional unit (circuit).
Up until now, most explanation techniques have taken a neuron-centric approach based on the idea that “a specific neuron detects a specific concept.” However, in reality, deep learning models form concepts through cooperative circuit structures involving many neurons. Based on this observation, the KAIST research team proposed a technique that expands the unit of concept representation from “neuron → circuit.”
The research team’s newly developed technology, Granular Concept Circuits (GCC), is a novel method that analyzes and visualizes how an image-classification model internally forms concepts at the circuit level.
GCC automatically traces circuits by computing Neuron Sensitivity and Semantic Flow. Neuron Sensitivity indicates how strongly a neuron responds to a particular feature, while Semantic Flow measures how strongly that feature is passed on to the next concept. Using these metrics, the system can visualize, step-by-step, how basic features such as color and texture are assembled into higher-level concepts.
The team conducted experiments in which specific circuits were temporarily disabled (ablation). As a result, when the circuit responsible for a concept was deactivated, the AI’s predictions actually changed.
In other words, the experiment directly demonstrated that the corresponding circuit indeed performs the function of recognizing that concept.
This study is regarded as the first to reveal, at a fine-grained circuit level, the actual structural process by which concepts are formed inside complex deep learning models. Through this, the research suggests practical applicability across the entire explainable AI (XAI) domain—including strengthening transparency in AI decision-making, analyzing the causes of misclassification, detecting bias, improving model debugging and architecture, and enhancing safety and accountability.
The research team stated, “This technology shows the concept structures that AI forms internally in a way that humans can understand,” adding that “this study provides a scientific starting point for researching how AI thinks.”
Professor Jaesik Choi emphasized, “Unlike previous approaches that simplified complex models for explanation, this is the first approach to precisely interpret the model’s interior at the level of fine-grained circuits,” and added, “We demonstrated that the concepts learned by AI can be automatically traced and visualized.”
This research was supported by the Ministry of Science and ICT and the Institute for Information & Communications Technology Planning & Evaluation (IITP) under the “Development of Artificial Intelligence Technology for Personalized Plug-and-Play Explanation and Verification of Explanation” project, the AI Research Hub Project, and the KAIST AI Graduate School Program, and was carried out with support from the Defense Acquisition Program Administration (DAPA) and the Agency for Defense Development (ADD) at the KAIST Center for Applied Research in Artificial Intelligence.
In an increasingly acute surgeon shortage, artificial intelligence could help fill the gap, coaching medical students as they practice surgical techniques.
A new tool, developed at Johns Hopkins University and trained on videos of expert surgeons at work, offers students real-time personalized advice as they practice suturing. Initial trials suggest AI can be a powerful substitute teacher for more experienced students.
In an increasingly acute surgeon shortage, artificial intelligence could help fill the gap, coaching medical students as they practice surgical techniques.
A new tool, trained on videos of expert surgeons at work, offers students real-time personalized advice as they practice suturing. Initial trials suggest AI can be a powerful substitute teacher for more experienced students.
“We’re at a pivotal time. The provider shortage is ever increasing and we need to find new ways to provide more and better opportunities for practice. Right now, an attending surgeon who already is short on time needs to come in and watch students practice, and rate them, and give them detailed feedback—that just doesn’t scale,” said senior author Mathias Unberath, an expert in AI assisted medicine who focuses on how people interact with AI. “The next best thing might be our explainable AI that shows students how their work deviates from expert surgeons.”
Currently many medical students watch videos of experts performing surgery and try to imitate what they see. There are even existing AI models that will rate students, but according to Unberath they fall short because they don’t tell students what they’re doing right or wrong.
“These models can tell you if you have high or low skill, but they struggle with telling you why,” he said. “If we want to enable meaningful self-training, we need to help learners understand what they need to focus on and why.”
The team’s model incorporates what’s known as “explainable AI,” an approach to AI that – in this example – will rate how well a student closes a wound and then also tell them precisely how to improve.
The team trained their model by tracking the hand movements of expert surgeons as they closed incisions. When students try the same task, the AI texts them immediately to tell them how they compared to an expert and how to refine their technique.
“Learners want someone to tell them objectively how they did,” said first author Catalina Gomez, a Johns Hopkins PhD student in computer science. “We can calculate their performance before and after the intervention and see if they are moving closer to expert practice.”
The team performed a first-of-its-kind study to see if students learned better from the AI or by watching videos. They randomly assigned 12 medical students with suturing experience to train with one of the two methods.
All participants practiced closing an incision with stitches. Some got immediate AI feedback while others tried to compare what they did to a surgeon in a video. Then everyone tried suturing again.
Compared to students who watched videos, some students coached by AI, those with more experience, learned much faster.
“In some individuals the AI feedback has a big effect,” Unberath said. “Beginner students still struggled with the task but students with a solid foundation in surgery, who are at the point where they can incorporate the advice, it had a great impact.”
Next the team plans to refine the model to make it easier to use. They hope to eventually create a version that students could use at home.
“We’d like to offer computer vision and AI technology that allows someone to practice in the comfort of their home with a suturing kit and a smart phone,” Unberath said. “This will help us scale up training in the medical fields. It’s really about how can we use this technology to solve problems.”
Authors include Lalithkumar Seenivasan, Xinrui Zou; Jeewoo Yoon; Sirui Chu; Ariel Leon; Patrick Kramer; Yu-Chun Ku; Jose L. Porras; and Masaru Ishii, all of Johns Hopkins, and Alejandro Martin-Gomez of University of Arkansas.
The team trained their model by tracking the hand movements of expert surgeons as they closed incisions.
The team trained their model by tracking the hand movements of expert surgeons as they closed incisions. When students try the same task, the AI texts them immediately to tell them how they compared to an expert and how to refine their technique.
Credit Johns Hopkins University
Can AI make us more creative? New study reveals surprising benefits of human-AI collaboration
Artificial intelligence (AI) is often seen as a tool to automate tasks and replace humans, but new research from Swansea University challenges this view, showing that AI can also act as a creative, engaging and inspiring partner.
A team from the University’s Computer Science Department has conducted one of the largest studies to date on how humans collaborate with AI during design tasks. More than 800 participants took part in an online experiment using an AI-powered system that supported users as they designed virtual cars.
Unlike many AI tools that optimise solutions behind the scenes, this system used a technique called MAP-Elites to generate diverse visual design galleries. These galleries included a wide range of potential car designs, including high-performing examples, unusual ideas and some deliberately imperfect ones.
Turing Fellow Dr Sean Walton, Associate Professor of Computer Science and lead author of the study, explained: “People often think of AI as something that speeds up tasks or improves efficiency, but our findings suggest something far more interesting. When people were shown AI-generated design suggestions, they spent more time on the task, produced better designs and felt more involved. It was not just about efficiency. It was about creativity and collaboration.”
A key insight from the study, published in the ACM journal Transactions on Interactive Intelligent Systems, is that traditional ways of evaluating AI design tools may be too narrow. Metrics such as how often users click or copy suggestions fail to capture the emotional, cognitive and behavioural dimensions of engagement. The Swansea team argues for more holistic evaluation methods that consider how AI systems influence how people feel, think and explore.
Dr Walton added: “Our study highlights the importance of diversity in AI output. Participants responded most positively to galleries that included a wide variety of ideas, including bad ones! These helped them move beyond their initial assumptions and explore a broader design space. This structured diversity prevented early fixation and encouraged creative risk-taking.
“As AI becomes increasingly embedded in creative fields, from engineering and architecture to music and game design, understanding how humans and intelligent systems work together is essential. As the technology evolves, the question is not only what AI can do but how it can help us think, create and collaborate more effectively.”
Researchers from Saarland University and the Max Planck Institute for Software Systems have, for the first time, shown that the reactions of humans and large language models (LLMs) to complex or misleading program code significantly align, by comparing brain activity of study participants with model uncertainty. Building on this, the team developed a data-driven method to automatically detect such confusing areas in code — a promising step toward better AI assistants for software development.
The team led by Sven Apel, Professor of Software Engineering at Saarland University, and Dr. Mariya Toneva, researcher at the Max Planck Institute for Software Systems, investigated how humans and large language models respond to confusing program code. The characteristics of such code, known as atoms of confusion, are well studied: They are short, syntactically correct programming patterns that are misleading for humans and can throw even experienced developers off track.
To find out whether LLMs and humans “think” about the same stumbling blocks, the research team used an interdisciplinary approach: On the one hand, they used data from an earlier study by Apel and colleagues, in which participants read confusing and clean code variants while their brain activity and attention were measured using electroencephalography (EEG) and eye tracking. On the other hand, they analyzed the “confusion” or model uncertainty of LLMs using so-called perplexity values. Perplexity is an established metric for evaluating language models by quantifying their uncertainty in predicting sequences of text tokens based on their probability.
The result: Wherever humans got stuck on code, the LLM also showed increased perplexity. EEG signals from participants—especially the so-called late frontal positivity, which in language research is associated with unexpected sentence endings—rose precisely where the language model’s uncertainty spiked. “We were astounded that the peaks in brain activity and model uncertainty showed significant correlations,” says Youssef Abdelsalam, who was advised by Toneva and Apel and was instrumental in conducting the study as part of his doctoral studies.
Based on this similarity, the researchers developed a data-driven method that automatically detects and highlights unclear parts of code. In more than 60 percent of cases, the algorithm successfully identified known, manually annotated confusing patterns in the test code and even discovered more than 150 new, previously unrecognized patterns that also coincided with increased brain activity.
“With this work, we are taking a step toward a better understanding of the alignment between humans and machines,” says Max Planck researcher Mariya Toneva. “If we know when and why LLMs and humans stumble in the same places, we can develop tools that make code more understandable and significantly improve human–AI collaboration,” adds Professor Sven Apel.
Through their project, the researchers are building a bridge between neuroscience, software engineering, and artificial intelligence. The study, currently published as a preprint, was accepted for publication at the International Conference on Software Engineering (ICSE), one of the world’s leading conferences in the field of software development. The conference will take place in Rio de Janeiro in April 2026. The authors of the study are: Youssef Abdelsalam, Norman Peitek, Anna-Maria Maurer, Mariya Toneva, and Sven Apel.
Preprint:
Y. Abdelsalam, N. Peitek, A.-M. Maurer, M. Toneva, S. Apel (2025): “How do Humans and LLMs Process Confusing Code?” arXiv:2508.18547v1 [cs.SE], August 25, 2025. https://arxiv.org/abs/2508.18547
Large language models (LLMs) such as ChatGPT and Gemini were originally designed to work with text only. Today, they have evolved into systems that can work with many types of information at once (multimodal systems), as well as understand and generate images, audio, speech and music.
The most common way to add speech to multimodal models is to convert it into small building blocks called audio tokens, which function for audio much like characters do for text. However, audio tokens still carry a lot of information, which makes speech harder to handle than text. Despite recent progress, integrating speech into large language models remains a major challenge.
“Because of this complexity, standard audio tokens often have a high bitrate (the amount of information packed into each second of audio). They pack a huge amount of information into each second of audio, which makes it difficult for large language models to learn from speech efficiently.”
A focus on speech’s meaning
Della Libera and his collaborators developed FocalCodec, a new audio tokenization method that compresses speech far more efficiently than previous approaches. It preserves both the sound and meaning of words at an ultra-low bitrate.
Instead of relying on heavy processing steps, the system uses a simple way of turning audio into compact units (binary spherical quantization) and a technique that helps the model focus on the most meaningful parts of speech (focal modulation). This makes the analysis faster and keeps the essential qualities of the voice intact.
To test FocalCodec, the team conducted a listening study with 33 participants who compared different audio samples. Participants often judged the reconstructed speech as nearly identical to the original recordings. This shows that the system can shrink speech significantly without making it sound robotic or distorted.
Recognized at a top AI conference
The work has been accepted at the Thirty-Ninth Annual Conference on Neural Information Processing Systems, one of the most selective conferences in machine learning and artificial intelligence.
“This work is particularly important, as it introduces a novel approach that can be highly valuable for building modern multimodal LLMs,” says Mirco Ravanelli, assistant professor and Della Libera’s supervisor. “By making speech lighter and easier to integrate, we move closer to AI systems that understand sound with the same confidence they bring to text.”
The paper also includes contributions from Francesco Paissan, visiting researcher at Mila and undergraduate student at the University of Trento, and Cem Subakan, affiliate assistant professor at Concordia.
Videos posted on social media by the Rusich paramilitary group feature pagan rituals, heavy machine guns and Nazi salutes and promote violence and supremacist ideology. © Telegram / dshrg2 - Molfar.Institute
This message was posted on Rusich Telegram channel on November 15. It calls on followers to send images of Ukrainian soldiers who have been executed in exchange for payment. © Telegram / dshrg2
