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)
Hegseth Brags About Maduro Abduction and Signs Military Deal on Secretive Venezuela Trip
“This visit is also a PR opportunity for Hegseth following the impeachment attempt and ahead of his congressional appearance,” said one former diplomat, anticipating “a US military base in Caracas soon.”
US Defense Secretary Pete Hegseth arrives at CATAM Military Air Base in Bogota on November 18, 2026. Photo by Maria Jose Gonzalez / AFP via Getty Images
While Colombia’s right-wing government on Wednesday rejected joint military operations with the United States, US Defense Secretary Pete Hegseth signed a sweeping deal with authorities in Venezuela—and bragged about American forces invading that country in January to abduct Venezuelan President Nicolás Maduro and his wife.
“It’s a great honor to be here as the first United States secretary of war this century to set foot in Venezuela. The weight of this moment does not escape me. Just one year ago, my presence here, our presence here, would have been deemed impossible—almost unimaginable,” Hegseth said. “But it was made possible for one simple reason: President Donald J. Trump had the courage to do what no president before him was willing to do.”
“The Department of War, in support of US law enforcement, brought Maduro and Celia Flores to American justice,” he continued, using Trump’s preferred name for the Pentagon. The Venezuelan couple is being held at Brooklyn’s Metropolitan Detention Center ahead of a trial scheduled for June. Maduro has called himself a “prisoner of war,” and both he and Flores have pleaded not guilty to narco-terrorism charges. Last month, the US addedtorture charges.
“Operation Absolute Resolve showed that the United States will not allow narco-terrorists to threaten Americans, erode our borders, nor run rampant throughout the Western Hemisphere. That is the Donroe Doctrine,” Hegseth declared. He celebrated the “close cooperation” with Venezuela’s interim president, Delcy Rodríguez, since the abduction and “a historic oil deal... that is bringing energy dominance for the benefit of Americans and for Venezuelans for decades to come.”
Hegseth also laid out the broad strokes of the newly signed deal that reestablishes a military partnership between the United States and Venezuela: “Training together, partnering together, sharing intelligence, more defense equipment and capabilities together.”
As The Guardian reported: During her five-minute address, Rodríguez, whose oil-rich country is widely seen to have become a US protectorate after January’s invasion, made no mention of Maduro, who is behind bars in New York City.
But referring indirectly to his capture and the subsequent reconciliation with Washington, she told Hegseth: “Thanks to the political and geopolitical maturity of our governments, we have—in a very short period of time, just a few months—turned what were very difficult and very tough times for our people, into an extraordinary space for cooperation and mutual benefit.”
Rodríguez hailed the “historic” energy deal her government recently struck with Trump, who has called it “the biggest oil deal in world history” as well as “the spoils” of his “conquest” of Venezuela.
Wall Street Journal national security reporter Lara Seligman said: “Hegseth is in Venezuela—but the Pentagon refused to confirm or give out any information about his trip. I’ve never seen anything like this before.”
Zachary Cohen, who covers the Defense Department for CNN, replied to Seligman that “as a result, the Pentagon has largely punted coverage of Hegseth’s trip to Venezuelan state TV.”
New York Times foreign correspondent Anatoly Kurmanaev told Cohen: “This is the most bizarre part of this trip to me, even more than that it happened or that these countries are now officially allies. It really takes a LOT to cede comms initiative to Venezuelan state TV.”
Former US Marine fighter pilot and Democratic congressional candidate Amy McGrath asked: “Remember when Pete Hegseth promised us the most transparent Pentagon in history? Why are we learning about new US military agreements with Venezuela’s authoritarian government from Venezuelan state television?”
Imdat Oner, a former Turkish diplomat who served in the embassy in Caracas and is now a senior fellow at Florida International University’s Jack D. Gordon Institute for Public Policy, noted that “two years ago, this very week, Iran’s defense minister visited Venezuela and signed a military cooperation deal. Today, Maduro’s former VP, Delcy Rodríguez, welcomes the US defense secretary to deepen military ties. The same military officers now work with the US.”
Oner said that he expects “to see a US military base in Caracas soon,” and also pointed to Congress’ recent unsuccessful effort to impeach Hegseth—whose department just failed its ninth straight audit while waging Trump’s illegal war on Iran.
“This visit is also a PR opportunity for Hegseth following the impeachment attempt and ahead of his congressional appearance,” Oner suggested. “As things in Iran gets further complicated, Venezuela will remain the Trump administration’s poster child for success.”
As for the administration’s rejection by Colombia on Wednesday, The New York Times reported that the “decision prompted the American defense secretary, Pete Hegseth, to alter the four-country trip he is now making to Latin America. Originally, he had been expected to meet on Wednesday with Colombian officials in Bogotá, the capital, to discuss the bilateral military agreement.”
“Instead, Mr. Hegseth was expected to touch down briefly to meet with Colombian officials after stopping in Caracas, Venezuela, where he met with the acting president and other officials,” the newspaper detailed. “The military agreement with Colombia, known as Title 10, would have enabled US forces to operate in that country and carry out joint airstrikes and ground operations.”
‘Shut It Down’: Vulnerable GOP Senator Pitches Abolishing Social Security During Fox Interview
“Social Security is the most popular and effective program in America. Ron Johnson wants to ‘shut it down.’ And any Republican who won’t commit to making billionaires pay their fair share agrees with him,” said Social Security Works. Sen. Ron Johnson (R-Wis.) speaks during a hearing of the Senate Homeland Security and Governmental Affairs Committee on February 12, 2026 in Washington, DC. (Photo by Chip Somodevilla/Getty Images)
A Republican senator up for reelection in a swing state in 2028 floated abolishing America’s most popular social insurance program during a Wednesday interview on Fox Business.
While appearing on the cable news network, Sen. Ron Johnson (R-Wis.) made his case for doing away with Social Security entirely.
“[Social Security] is a scam, it is grotesque,” he said. “At this point, there’s really no way to fix Social Security. It’s a Ponzi scheme, you can’t fix it. You just shut it down and hope that you can recovery some of your money.”
The Wisconsin Republican’s attack on Social Security came just hours before he was set to preside over a Senate Budget Committee hearing titled “Social Security: The Reality,” which Johnson claimed would demonstrate how the program has been “grossly mismanaged.”
“Social Security is the most popular and effective program in America,” the group wrote in a social media post. “Ron Johnson wants to ‘shut it down.’ And any Republicans who won’t commit to making billionaires pay their fair share agrees with him.”
Nancy Altman, president of Social Security Works, described Johnson’s call to end Social Security as “a brazen effort, in broad daylight, to steal Americans’ earned benefits.”
“Sen. Johnson has an estimated net worth of $40 million,” Altman added. “He wants to shut Social Security down so that he never has to contribute his fair share. If he gets his wish, he will still enjoy a secure retirement. Meanwhile, millions of seniors and people with disabilities will face homelessness and hunger.”
Stephen Nuñez, director of stratification economics at the Roosevelt Institute, wrote that Johnson’s claim about there being “no way to fix Social Security” is “complete horseshit,” pointing to an analysis he published earlier this year explaining the real reasons for the program’s projected fiscal shortfall.
“Social Security’s current difficulties are a combination of the quirks of the laws governing it (a separate and limited funding source, no capacity to borrow, etc.),” Nuñez explained, “combined with some important ‘misses’ in actuarial projections and legislators’ unwillingness to legislate.”
“There is no bankruptcy or collapse in the cards,” Nuñez emphasized. “Just like in 1981, we are simply waiting for lawmakers to act.”
Economist Matt Darling, a policy fellow at Searchlight Institute, similarly pushed back on Johnson’s fearmongering about Social Security’s finances.
“Regular reminder that the worst-case scenario for Social Security is something like ‘pays out 80% of benefits,’” explained Darling, who noted that this would still make Social Security “roughly the equivalent of the old age program in Japan.”
Independent policy analyst Ned Resnikoff predicted that Johnson’s attacks on Social Security will end poorly for him when he runs for reelection in two years.
“The last time they tried to privatize Social Security was when they were flush with victory after the 2004 election and it brought their political momentum to a dead halt,” Resnikoff observed, referring to former President George W. Bush’s failed privatization effort.
“Bringing it up again after a historic rout is just evidence of a death wish.”
Senate Democrats Pile On GOP’s Ron Johnson After He Floats Abolishing Social Security
“This is the first time in decades that a senior senator... said he would rather shut down Social Security than fix it,” said US Sen. Ron Wyden.
Democrats on the US SenateBudget Committee on Wednesday took turns slamming a Republican colleague who floated abolishing Social Security.
Earlier in the day, Sen. Ron Johnson (R-Wis.) appeared on Fox Business, where he described Social Security as a “scam” that cannot be reformed.
“At this point, there’s really no way to fix Social Security,” Johnson said. “It’s a Ponzi scheme, you can’t fix it. You just shut it down and hope that you can recovery some of your money.”
Democratic senators wasted no time in pouncing on Johnson’s claims about the country’s most popular social insurance program during Wednesday’s hearing.
Sen. Jeff Merkley (D-Ore.), ranking member of the Senate Budget Committee who is in line to be its chairman next year, accused Johnson of peddling baseless fears about Social Security going bankrupt.
“I think the biggest factor scaring Americans right now was your quote this morning,” Merkley said. “That does terrify people across the country. And much of what my colleague has described is the fact that he does not like the way that Social Security is structured.”
Merkley also tore into Johnson’s claim that there is no way to fix Social Security’s finances when a significant portion of it can be solved by raising or even eliminating the cap on wages subject to the payroll tax.
“This year, I held 36 town halls, one in each county in Oregon,” Merkley said. “And in each one, we did a survey on how people would like to see Social Security addressed to make sure it is solvent. The results were clear and consistent... from cities to rural communities. The top response from citizens was, ‘Raise the cap on the income to which premiums apply.’”
“Most think it’s unfair,” the Oregon Democrat added, “for a local firefighter who pays the exact same contribution as a billionaire hedge fund manager.”
Sen. Ron Wyden (D-Ore.) similarly denounced Johnson for wanting to scrap Social Security rather than make the ultrarich pay higher taxes.
“This is the first time in decades that a senior senator, chairman of the Budget Committee, said he would rather shut down Social Security than fix it,” Wyden said, “in effect, saying he would rather throw in the towel and have billionaires pay their fair share. That is, in my view, an astounding statement.”
Sen. Mark Warner (D-Va.) told Johnson that he was “shocked and disappointed” to see Johnson’s call to shut Social Security down, to which the Wisconsin Republican replied that he was being “taken out of context.”
“No,” Warner shot back. “I’m reading a quote from you on Fox News!... We don’t want to scare the American public. This is outrageous scare tactics. The American people deserve better.”
Johnson, who is up for reelection in 2028, is projected as one of Democrats’ top targets as they aimed to further expand their majority in the US Senate.
The Progressives Out Campaigned the Far-Right; Now Comes the Challenge of Leadership
To deliver for voters, Democratic lawmakers will need to change their leadership, challenge neoliberal orthodoxy to fight inflation, and hold Trump accountable for his crimes. Congressional candidate Claire Valdez, Congressional candidate Brad Lander, New York Mayor Zohran Mamdani, and Congressional candidate Darializa Avila Chevalier raise their hands during a Get Out the Vote (GOTV) rally at King’s Theater on June 18, 2026 in New York City. (Photo by Michael M. Santiago/Getty Images)
People who are not American citizens like me don’t have a vote in the US elections, yet what happens at the center of the empire has an impact, positive or negative, on us. So for us, analyzing what happened on November 3 and what its implications and consequences are is an urgent, necessary exercise.
This analysis is divided into three parts: what the far-right did wrong, what progressives did right, and what are the challenges confronting US progressives in the run-up to the 2028 presidential elections.
What the Far-Right Did Wrong
Where did the far-right go wrong? Here I found two books that I’ve just read particularly illuminating: David Ost’s Red Pill Politics and Naomi Klein and Astra Taylor’s End Times Fascism. One of the key points both books put forward is that the far-right can’t rule via terror alone. It needs to have a positive program that brings economic and other benefits to their base and the groups they want to attract.
Snowy Minneapolis was the turning point of the struggle against 21st century American fascism, much like freezing Stalingrad was the graveyard of the German Sixth Army during the Second World War.
Even as they terrorized socialists, Jews, and Roma people, fascists in the early 20th century in Germany and Italy offered social security measures, employment benefits, housing programs, massive construction projects, and leisure opportunities to their mass base of workers and middle-class people.
President Donald Trump over the last two years, in contrast, has ruled purely negatively, terrorizing migrants and people of color, raising tariffs to punitive levels, engaging in a senseless war. He didn’t offer any economic benefits to his white base, perhaps thinking that they would be content with the psychological satisfaction of seeing Immigration and Citizenship Enforcement (ICE) agents cruelly arrest and deport migrants and shooting anyone who resisted, even if they were white people.
But lack of a positive program for the white base was not the only thing that stymied the far-right. There was just the total chaos of Trump’s reign—allowing Israeli leader Benjamin Netanyahu to enmesh him in an endless war with Iran after he had promised to be the “peace president,” denying that the United States had lost the war when it had clearly done so, denying the ravages of inflation on people’s pocketbooks. The far-right base does not mind the breaking of immigrants’ bones, but they cannot stand perpetual chaos and instability. That’s why many either crossed the political line and voted for the Democrats on November 3 or simply decided not to show up at the polls. The problem the American far-right faces is that their chosen leader is a narcissist first and only a fascist second.
What Progressives Did Right
What did the progressives do right? Here I would highlight two things: the heroic resistance of ordinary people to ICE in Minnesota and the candidacy of Zohran Mamdani in New York City.
The impact of what happened in Minnesota in the early months of this year cannot be underestimated. The sight of masses of people coming together to protect their neighbors from being kidnapped by ICE agents invading their community provided inspiration not just to progressives but to ordinary decent people throughout the United States. If Americans did not want the authoritarian rule and cruelty the far-right wanted to impose on the country, they had to go out to the streets to defend their rights.
Naomi Klein and Astra Taylor say it best: In the snow and tear gas of Minnesota’s winter, the far-right lost control over its central story about who is replacing whom and who is invading and who invaded… When thousands of masked and armed agents from out of town descended on an American city, that narrative could not hold. As residents witnessed strangers in tear gas and kidnap their parents, a profound moral revulsion set in. Huge numbers of people many of whom had never protested before, knew something deep in their bones: they were literally being invaded by an outside force. And they were bound and determined not to let their diverse and loving community be replaced by a police state.
Snowy Minneapolis was the turning point of the struggle against 21st century American fascism, much like freezing Stalingrad was the graveyard of the German Sixth Army during the Second World War.
The second key development that propelled the Democratic wave was Zohran Mamdani’s campaign for mayor of New York in 2025. Mamdani came up with what became the playbook for the 2026 elections during his campaign for mayor of New York City. There were three key elements:Don’t give too much credence to the polls, run and run hard, and the numbers will follow; Run an inclusive campaign, spreading hope and unity instead of sowing division and hate; and Run on affordability and don’t let the other side derail you with false issues, like accusing you of being antisemitic.
What began as a one-man campaign against the odds became at its end a massive collective effort. Again, Klein and Taylor say it best: The campaign with its tens of thousands of volunteers making phone calls and knocking on doors [helped] New Yorkers to connect with another in ways many had not previously. And while they found joy and community, they also came face-to-face with neighbors who were struggling. The people who made and delivered their food were being ripped off. The people who nursed them when they were sick and drove them in cabs and buses were barely scraping by, many families were at imminent risk of eviction. he deepening realization—in a city overflowing with private wealth—made the threats of an ultrarich exodus less frightening than enraging.
The Mamdani playbook was successfully reproduced by Democratic candidates in state after state. In fact, it was adopted even beyond the borders of the US—for instance, in Berlin, where the progressive Die Linke party got the highest percentage of votes during the German regional elections last September. Challenges to the Left
Let’s now turn to the challenges that confront progressives in the run-up to the presidential elections of 2028.
For the left, the first challenge will be how to make sure that replacing their leadership will not result in intra-party bloodletting that will distract them from pushing their affordability programs. It’s fairly clear that Sen. Chuck Schumer (D-NY) and Rep. Hakim Jeffries (D-NY) are not the people who can lead the Democrats at a time that people are expecting an aggressive approach to their multitude of problems, from dealing with racism and ending the attack on migrants to stopping inflation and bringing Trump to justice. Schumer and Jeffries are classic compromisers, and Trump will try to corner them into compromises that will blunt the fulfillment of the Democrats’ promises in the lead-up to 2028 and allow the Republicans to regain lost ground.
So, the big challenge to the Democrats is how to avoid a debilitating internecine struggle and come up with the equivalent of a golden handshake that will allow Schumer and Jeffries to retire gracefully from the competition for leadership.
In short, solving the affordability question necessarily moves the left closer to socialist solutions, so the Democratic Socialists of America and other progressives should expect full-blown resistance from the capitalist establishment and should be prepared for a full-scale ideological battle to win over the allegiance of the people.
The second challenge is how to meet the very real promises of affordability that progressives ran on. Most pressing is how to bring inflation under control. There’s no way to do this except through controls on prices, and here they’ll come into conflict with the fundamental tenet of neoliberal ideology that you can’t interfere with the market. To solve the crisis of housing you can’t trust private developers, so government must manage either directly or indirectly the rapid construction of tens of millions of housing units. Government must move to decisively ban the construction of data centers and radically regulate AI, and this will bring progressives into sharp conflict with Silicon Valley and the billionaire class, as will taxes on the latter’s wealth.
In short, solving the affordability question necessarily moves the left closer to socialist solutions, so the Democratic Socialists of America and other progressives should expect full-blown resistance from the capitalist establishment and should be prepared for a full-scale ideological battle to win over the allegiance of the people. Even if the moderates do not want to frame it as a capitalism-versus-socialism battle, the conservatives and the far-right will do so.
The third challenge is making Trump accountable for his multiple transgressions of the law. Clearly, the Democratic base wants this reckoning and will not allow him to go scot-free. But it will not be easy to bring Trump to heel, since he fights best when cornered. Also, being a narcissist, he has a Samson complex: If he’s going down anyway, he’ll want to bring everything down with him. So, Trump and the far-right must not be underestimated since they’re most dangerous when they’ve suffered a massive defeat. You’ll need not just determination and courage but a smart strategy. And that’s one reason the struggle for the Democratic Party leadership is such a high-stakes battle.
Walden Bello Walden Bello is the co-founder and current senior analyst of the Bangkok-based Focus on the Global South and the International Adjunct Professor of Sociology at the State University of New York at Binghamton. He received the Right Livelihood Award, also known as the Alternative Nobel Prize, in 2003, and was named Outstanding Public Scholar of the International Studies Association in 2008. His books include: "Counterrevolution: The Global Rise of the Far Right" (2019) and "Capitalism's Last Stand?: Deglobalization in the Age of Austerity" (2013). Full Bio >
Wednesday, November 18, 2026
A third of the global AI supply chain is ‘invisible’ to governments, study finds
Researchers identify 137 supply chain elements for global AI production. They compare these to government documents from dozens of countries and find 33% of the elements, mainly from the “physical middle” of AI production (such as wafers, packaging and networking), are largely absent from top-level AI policy worldwide.
“AI sovereignty” is an impossibility for any one nation, including the US and China, and countries such as Japan and the UK have far more influence over the supply of AI than they realise, say experts. A country-by-country report is available alongside the main report.
As much as a third of the AI supply chain receives no attention from national and international policymaking and regulation, according to a new Cambridge report that reveals a vast “zone of invisibility” in the middle of global AI production.
A University of Cambridge team found vulnerabilities across 14 supply-chain layers, with 137 elements produced or convened in over fifty countries. Many critical “chokepoints” sit outside the AI powerhouses of China and the US.
“The US and China are central to the AI story, but this research shows us the extent to which critical technologies and materials across the AI stack have dependencies that extend far beyond their borders,” said Verity Harding, Director of the AI and Geopolitics project. “It shows that no one country can control the AI supply chain alone.”
One Japanese firm has a 95% market share of a material needed for AI chips, the report shows. Even notorious chokepoints depend on other nations: the Dutch firm dominating EUV lithography (ASML) relies on essential parts controlled by German and US companies.
In fact, the study suggests that countries such as Japan, Germany, and some so-called middle powers – from Belgium to South Korea – have far more leverage over global AI production than even their own governments realise.
Researchers describe an “unexpected mismatch” between a nation’s top AI policy documents and its actual positions in the supply chain, calling it a “universal pattern”.
“Most policy discussions reduce artificial intelligence to just a few inputs, such as compute, data and models. This is not an accurate understanding of the complexity of the AI supply chain,” said lead author Dr Siyi Liu.
“Many governments are unaware of the strengths and vulnerabilities they hold at the heart of the global AI supply chain. Understanding this will be the basis of credible bargaining power in AI governance.” she added.
The UK is involved in 41 of the 137 supply-chain elements researchers identify, yet its AI policies name just three, ignoring UK specialisms in chip design and industrial gases. Japan is second only to the US in terms of supply-chain positions, yet its AI policy documents recognise only three.
Cambridge researchers analysed 117 AI-related policy documents from 43 countries and territories, the EU and OECD, and found a fixation on models and datacentres, as well as funding and governance, while the factories, chemicals, chip-design software and packaging technologies vital to the AI supply chain are completely overlooked.
Disruption to any one of these “upstream” components could trigger a domino effect that leads to AI supply shortages, potentially destabilising stock markets, investment and productivity around the world.
Experts from the AI & Geopolitics Project (AIxGEO) at Cambridge’s Bennett School of Public Policy say that total “AI sovereignty”, a nation autonomously operating its own AI supply chain – a goal many nations hint at pursuing – is “impossible”.
The research team created a toolkit for governments to audit their place in the chain. They say it will help policymakers understand where and when to diversify across AI element providers, who they depend on, or when to “exit” entirely from an outside supplier and build national capacity.
For the latest study, the AIxGEO team comprehensively mapped the global AI supply chain to reveal 14 distinct layers, from the raw minerals and the ultra-pure gases needed for semiconductors, to the wafers, chips and servers, and eventually the trained models. Cross-cutting layers include governance standards and shipping routes.
Breaking it down, researchers identified 137 specific “elements”, such as gallium or hyperscale cloud platforms, with a further 351 “sub-elements”: tech specs that determine who can supply it. Nearly 800 firms identified hold supply-chain positions across 56 countries and territories.
However, the team’s analysis of AI policy and regulation documents from dozens of countries as well as intergovernmental organisations shows that just over 33% of the identified supply-chain elements are simply not mentioned.
Most of this “invisibility” is in the “physical middle” of the chain. This includes chip design, packaging and the testing of chips, and some of the hardware and software behind “compute” such as making AI servers and “merchant accelerators” (platforms that allow businesses to adopt AI).
An example of a hidden chokepoint highlighted by the report is the EDA (Electronic Design Automation) software used to design AI chips. It is essential for companies like NVIDIA, AMD and Google.
This software is dominated by three firms – Synopsys, Cadence and Siemens EDA in the US – yet receives almost no policy attention, unlike the chips and LLMs that wouldn’t exist without it.
Another major “blind spot” is advanced packaging: connecting multiple semiconductor chips into high-performance AI processors essential for frontier models. This packaging is overwhelmingly concentrated in Asian countries and territories such as Taiwan, Vietnam and the Philippines.
“Policymakers focus on finished chips rather than the ecosystem behind them. They focus on GPUs but ignore the packaging operations needed to turn those chips into usable AI hardware,” said Liu.
Japan emerges as the world’s most underrated AI power, with the second largest number of supply chain positions, directly ahead of China. The report shows 101 Japanese firms hold positions across 69 elements, from silicon wafers to semiconductor testing. However, Japan’s own policy documents feature only three of these elements.
The researchers point out that a Japanese company, Ajinomoto, controls more than 95% of the global supply of build-up film, an insulating material needed for advanced AI chip manufacturing.
“Japan’s strength comes from equipment, and without it there's no way you can produce the hardware,” said Liu. “The supply chain is sequential, so the trickle down could affect computing capacities of the top frontier models.”
The UK has little AI-related manufacturing, but holds leverage in chip design IP through Cambridge company Arm (even with majority Japanese SoftBank ownership). It is a base for research and expertise, particularly in AI security, and home to valuable health and genomic datasets.
The report highlights absences from reviewed UK policy where it has footholds in the supply chain, such as chip design, and the pure gases needed for semiconductor plants via Linde, one of the world's largest industrial gas companies.
Some of the smaller nations with more clout than they might realise include Belgium, which is strong in recycled semiconductor chemicals and has the major nanoelectronics R&D centre Interuniversity Microelectronics Centre (imec).
Ireland has no production supply-chain position, yet it can affect AI capacity through energy and planning as it provides a European base for big US tech e.g. Google and AMD, making it a “host and infrastructure state”. While Qatar produces much of the world’s helium, a vital supply-chain raw material, and is also a huge potential AI investor.
Earlier this year, Nvidia’s former-CEO Jensen Huang described the AI supply chain as a “five-layer cake” (energy, chips, infrastructure, models, and applications).
Verity Harding added: “The AI supply chain is far more complex, and power over it far more diffused, than some simplified narratives might have us believe. Many governments hold more cards than they realise.”
Notes:
The 14 supply chain layers identified by the report: minerals that must be scooped and refined; ultra-pure gases and chemicals required by a fabrication plant or facility; wafers; machines that print the circuits; chips; memory and packaging around chips; servers; data centres; power and cooling that run data centres; cloud platforms; data; models.
About the AI & Geopolitics project (AIxGEO)
AIxGEO, led by Verity Harding, exists to interrogate the framing of AI development as a zero-sum arms race between two superpowers, and to shift the policy debate towards a more cooperative future so the technology can deliver the broad benefits it promises.
About the Bennett School of Public Policy
The Bennett School is a leading school of public policy in the UK and wider world, for public policy thinking, research, teaching and engagement with practitioners from policy, business and civic organisations. Drawing on Cambridge’s broad talent base, it supports effective decision-making through interdisciplinary research, training, and collaboration.
Pulling on an AI model reveals how it represents physical relationships, and how we can train better AI models that understand physics from first principles
Dongguan Institute of Materials Science and Technology, CAS
A spring illustrates the mechanics-inspired idea: a system’s behavior is revealed by observing how it responds to a controlled change. In the language model, paired scientific questions produce changes in internal representations, depicted as interconnected networks. Targeted interventions then test how selected representations influence the model’s answer preferences. The figure is a conceptual illustration, not a visualization of measured data. Image credit: Markus J. Buehler/MIT.
Large language models are increasingly used to explore scientific knowledge, analyze materials and generate ideas. Yet a correct answer alone does not reveal how a model arrived at it. The answer might reflect a physical relationship, a familiar phrase or a statistical shortcut.
Physical laws describe relationships between quantities. An increase in one quantity can increase a material's response under one law and decrease it under another. Testing whether an AI model represents that distinction requires more than examining its response to a single question.
The Solution
Markus J. Buehler, the Jerry McAfee Professor of Engineering at MIT, investigated this problem using three open-weight models from Google's Gemma family. Inspired by mechanics, he compared matched questions in which a physical condition changed while the surrounding context was controlled. He then measured how the model's hidden states, the internal numerical representations it creates while processing a question, moved in response.
Across 60 materials-science laws, the resulting changes tracked the ordering of direct, neutral and inverse relationships with a rank correlation of 0.910. They correctly oriented 39 of 40 directional laws. When equations and prescribed reasoning instructions were removed, the models still oriented 38 of 40 laws correctly at the smallest tested scale and all 40 at the two larger scales.
The study also tested whether scientific concepts could be read from intermediate hidden states when their names were omitted from the prompt. Some concepts remained recoverable, while others did not. This established a distinction between information that can be read from a model and information that is used in a decision.
To test causal use, the author modified selected internal representations without retraining the model or adding verbal hints. In one controlled test, a fixed direction associated with grain size shifted answer preferences appropriately across 12 grain-refinement and grain-coarsening cases. Further interventions across 24 relationships showed selective causal contributions across all three model sizes.
The effects were context-dependent: the same grain-size intervention did not transfer successfully to every new wording or physical regime tested. The experiments therefore distinguish three levels of evidence: readability, representation and causal use.
The Future
The next step is representation-aware scientific AI: training and evaluating models not only on their answers, but also on whether their internal representations respond consistently to changes in physical conditions. This opens a path to testing models against scientific relationships and identifying where they rely on shortcuts.
The present work establishes the measurements and interventions for selected, known relationships in controlled tasks. Extending them to new scientific problems and model architectures is the next challenge, including applications in developing enhanced training rewards for AI models that foster broader and deeper generalizations of physics principles. This can lead to new first-principles based AI models that think and reason from foundational concepts and that may be far more efficient and powerful than current models.
The Impact
By treating an AI model's internal representations as objects that can be measured and experimentally altered, this research brings an approach inspired by the physical sciences to the study of AI itself. It gives researchers a more precise way to ask what a model represents, what it uses and where its scientific reasoning breaks down.
Publication: M.J. Buehler, "Reading and Steering Representations of Materials-Science Mechanisms in an Open-Weight Language Model," AI for Science2, 045001 (2026). https://doi.org/10.1088/3050-287X/aeb27b
Why do some artworks stick in our minds, while others fade away soon after we view them? A new study published in Psychological Science suggests that it’s partly because the artists themselves can imbue memorability into their works.
Researchers at the University of Chicago tested this ability by conducting a nationwide art contest, asking artists across the country to create works that were intended to be either memorable or forgettable. They found that not only were artists successful in completing this task, but that a deep learning model was most adept at predicting memorability. These findings provide insights into how artists may intuitively understand memory and how technology can predict it.
“Our study suggests that artists can offer valuable insights into memory,” said Yifei Chen, a coauthor of the new study. “They seem to have an intuitive sense of what makes artworks actually stick in people’s minds.”
Chen and her colleagues collected art contest submissions over the course of three months for the study. They received a total of 86 submissions that adhered to the contest’s rules and first tested the memorability of the artworks with a group of online participants. Overall, the researchers found that the works submitted to the “memorable” category were more likely to be remembered, and that “forgettable” works tended to slip the mind. Analyses also showed that simpler artworks, as well as ones with fewer repetitive elements and less symmetry, had more memorability.
The team then chose 20 of both the most memorable and most forgettable artworks from these subsets. They asked the artists to ship their art to Chicago for a month-long exhibit in an in-person gallery. When the exhibition opened, visitors could freely move through the gallery, after which their memory of the artworks was tested. “The same pattern held in the gallery: Artworks that were more memorable online were also better remembered in person,” Chen said.
“It was a very risky experiment for our lab,” noted Wilma Bainbridge, senior author of the paper and Chen’s advisor, as the gallery only had 40 pieces and had an overarching theme of memory. The process of the experiment was also captured in a short documentary.
“But even still, they forgot the forgettable pieces and remembered the memorable ones, which shows how strong this effect really is,” she said. “It works on the order of just a few minutes, even in highly motivated people in this real-world setting.”
Finally, the researchers tested if memory of artworks could be predicted by artists, online laypeople without any art expertise, or a neural network model called ResMem that was developed by researchers to predict image memorability. In the new study, ResMem performed best at distinguishing the most and least memorable artworks, followed by artists’ submission categories, then ratings by laypeople.
“ResMem had only been trained on photographs,” Chen explained, making these new, original works a test of its ability. And yet, it did “much better than any human can do,” she said, highlighting that “what people think is memorable is different from what actually is.”
Although artists weren’t as proficient as AI in predicting memorability, their success in designing for memory suggests they could be valuable scientific collaborators. Studying their creative choices could help researchers understand which visual features support memory and how those features might be deliberately shaped. “This gives us so much potential for tapping into what drives memorability,” Chen said.
Refining models like ResMem could also help artists find tangible ways to make their works stand out or aid exhibit designers with how to arrange galleries. Bainbridge adds that ResMem could also be applied in a broader educational context, such as finding out what kinds of visuals may help children, patients, or other groups learn information that is difficult to remember. “Our preliminary findings are that if you have a memorable picture with all this information, they learn that additional information better, too,” Bainbridge said.
These findings contribute to a growing body of work examining how the properties of images and the real-world contexts they are encountered in can shape what we remember, Chen said. Because memory research has often focused on processes within the viewer, studying the relationships of visuals and memory in everyday settings is, as Chen says, “a direction that the whole field could explore further in the future.”
Researchers put AI's understanding of people to the test
The framework helps scientists determine whether AI can identify meaningful patterns in human experiences and shows where human insight makes a difference
University of Missouri researchers have developed a way to determine whether artificial intelligence is uncovering meaningful insights from people’s words or simply sorting them into categories.
From interviews and focus groups to open-ended survey responses, researchers look for patterns in what people say to better understand their experiences and perspectives. But identifying meaningful themes takes time, and interpreting people’s words requires more than simply grouping similar responses together.
To determine whether AI can identify meaningful patterns without losing important nuances, a Mizzou team spanning journalism, data science and nursing developed a method to evaluate computer-generated topics. The approach examines how closely those topics match themes identified by human researchers and whether they provide useful insights into the information.
“AI can help researchers see patterns across large amounts of information, but the real question is what those patterns tell us about people’s experiences,” said Chau Tong, an assistant professor in the Missouri School of Journalism with a joint appointment in the MU Institute for Data Science and Informatics. “By bringing together different areas of expertise, we can better evaluate how these tools support research and where human judgment remains essential.”
Putting AI’s analysis to the test
The team tested the evaluation framework, called the Alignment and Usefulness Rating Audit (AURA), by analyzing 32 interviews with nursing faculty, practitioners and students about their experiences with generative AI.
Researchers first identified key themes in the interviews, then compared their findings with topics generated by a computer model. AI identified many of the same broad themes as human researchers, but it often missed important distinctions in meaning and context.
Eleven of the 12 themes identified by human researchers closely matched at least one topic generated by the computer model. However, a match did not necessarily mean the model captured the full meaning of a theme or organized the information in the most useful way.
The model sometimes combined several distinct themes into one broad topic or divided a single theme into several smaller ones.
“When people talk in interviews, they don’t usually stay on one topic from start to finish,” said John Robert Bautista, an assistant professor in Mizzou’s Sinclair School of Nursing with a joint appointment in the MU Institute for Data Science and Informatics. “AI can help organize those conversations so researchers can more easily see what people are saying.”
Where AI helps and where it falls short
While AI helped researchers organize large amounts of information and identify patterns, the team found less evidence that it consistently generated meaningful new insights beyond what human analysts identified.
“Our findings suggest the technology can help researchers manage large collections of interviews, focus groups and survey responses,” said Temitope Eyinla, a doctoral researcher in the MU Institute for Data Science and Informatics. “But human expertise remains central to interpreting the results.”
The evaluation could support research into issues such as patient experiences, educational challenges and workplace concerns, where understanding people’s perspectives is essential. However, the study examined a relatively small set of interviews, so additional testing is needed to determine how well AURA works across different subject areas and types of research.
The European Laboratory for Artificial Intelligence (ELAI) has won the call by the European Commission to pilot a virtual institute of world-class excellence under the Resource for AI Science in Europe (RAISE). The announcement was made today in Brussels at the Apply AI Summit in the European Commission and Martin Brynskov, ELAI Coordinator from the University of Copenhagen.
ELAI brings together a consortium of 17 sites, 33 partners, and 21 supporters in 17 countries, including Japan and Canada, on 3 continents, to establish a distributed virtual institute, built on excellent research institutions creating the structure for leading researchers to tackle, at the European level, Frontier AI research challenges, including in efficiency, safety and security. With €17M seed funding over five years, it will amplify a strategic research agenda that pools and provisions resources for scientific breakthroughs to ensure European leadership in AI research and innovation to deliver on societal challenges.
'ELAI unites leading AI research institutions combining world-class fundamental AI research with deep expertise in machine learning for the sciences. Building on almost a decade of collaboration through the European Laboratory for Learning and Intelligent Systems (ELLIS), we have proven ways to organise joint research and attract internationally leading talent at a scale no single site or national effort could achieve alone,' says Serge Belongie, Professor at University of Copenhagen, Director of the Pioneer Centre for Artificial Intelligence, and Board President of ELLIS.
The announcement was made at the Apply AI Summit in Brussels, which marks one year since the adoption of the Apply AI Strategy and concludes the first European AI Innovation Month. The Summit showcases how trustworthy, human-centric and innovative AI can boost the competitiveness of sectors such as healthcare, mobility and the public sector, with European use cases already running on homegrown AI models and systems.
'Europe’s research strength is spread across many institutions and countries, and RAISE exists to bring it together. ELAI will do that for the science behind AI itself, uniting leading institutions around the questions that matter most for AI Europe can rely on. Stronger foundations in research are what allow the Apply AI Strategy to deliver, and what will help make Europe an AI Continent,' said Cécile Huet, Head of the Unit Robotics and Artificial Intelligence Innovation and Excellence in the European AI Office at the European Commission.
RAISE is a pilot launched by the European Commission to deliver on the AI in Science strategy, which supports and incentivises the development and use of AI by the European scientific community. It pools strategic resources (funding, compute, data and talent) and operates along two pillars: Science for AI, which supports basic research to advance core AI capabilities, in particular safe and secure frontier AI; and AI in Science, which promotes the use of AI for progress in different scientific disciplines. ELAI is funded under the Science for AI pillar.
ELAI’s activities will begin on 1 January 2027.
About ELAI The European Laboratory for Artificial Intelligence (ELAI) establishes a distributed, excellence-driven virtual institute that organises leading researchers around a shared frontier research agenda to advance science for AI, establishing new foundations for efficient, reliable and interactive learning systems. It is funded by the European Commission and led by the University of Copenhagen.
AI system developed to recommend content for improving emotional health
Imagen de la aplicación EMOTINET, que recomienda contenidos adaptados para mejorar su salud emocional. Aparece un chatbot que pregunta: María, selecciona los temas que te interesaría trabajar en Emotinet
Researchers at the Universidad Carlos III de Madrid (UC3M), in collaboration with the company Psiconnea, have developed an artificial intelligence (AI)-based system capable of detecting emotional patterns in individuals and recommending tailored content to improve their emotional well-being. The platform, available via the EMOTINET app, enables users to perform self-assessments to spot early signs of psychological distress and—through exercises, visualizations, and professionally guided series—overcome mental blocks, reduce anxiety, and gain emotional clarity.
The contribution of UC3M's Robotics Lab to this university-to-society knowledge transfer project consists of a recommendation system similar to those used by streaming platforms, adapted specifically to the mental health field. "Just as multimedia platforms recommend series and movies based on our preferences and viewing history, this system works in a similar way," explains José Carlos Castillo, a researcher in the Social Robotics group at UC3M's Robotics Lab.
To achieve this, the app developed by Psiconnea (a Spanish company of psychotechnologists specializing in emotional health) processes various types of data, including the user's psychological profile, their interaction with the platform, and their self-reported needs. The AI system designed by UC3M integrates all this information to generate personalized multimedia recommendations. Specifically, these consist of brief explanatory videos—acting as "micro-learning modules"—that help users better address their specific challenges and have been pre-validated by Psiconnea.
A First Step, Not a Substitute for a Professional
The researchers emphasize that the tool is not intended to replace mental health professionals. "It really serves as a preliminary step: when someone senses they have a problem, the app allows them to evaluate it and offers an initial approach. If the issue persists or cannot be resolved directly, the system facilitates referral to a specialist and access to a consultation," clarifies José Carlos Castillo.
EMOTINET also features emotional self-assessments, psychologist-guided exercises, and an anonymous support community. According to Psiconnea CEO Rosa Becerril, more than 80,000 people have used the platform, and 80% of users reduce their levels of stress, anxiety, or depression.
The UC3M team highlights the value of this type of collaboration. "We are deeply grateful for the company's commitment to university knowledge transfer, because for us, these synergies are key: universities hold vast amounts of knowledge that need to reach society, which is often no easy task. Through initiatives like this, we can translate research into something tangible that directly helps people," concludes José Carlos Castillo.
Seven research-based practices offer steps for evaluating and introducing AI tools with a focus on trust, human oversight, job design and workers’ well-being
WASHINGTON — As organizations introduce artificial intelligence tools to their workforce, people may be left to navigate unclear policies, conflicting expectations and new questions about trust and acceptable AI use in the workplace.
A new guide released by the American Psychological Association outlines what organizations should consider when introducing AI into people’s work.
Created for organizations seeking to adopt AI responsibly while also supporting employees and improving how work is designed, Deploying AI Into Workplaces: What Leaders Need to Know offers seven best practices based on established psychological research and emerging evidence about AI. These range from matching safeguards to the stakes of potential errors to ensuring that AI is introduced through a fair process.
“The rapidly growing use of AI at work is creating new behaviors and social judgments while raising questions about trust, all of which are fundamentally questions of psychology. Did you use AI to write your emails today? Am I supposed to think more or less of you if you did?” said APA CEO Arthur C. Evans Jr., Ph.D. “Psychological science can help organizations understand not just what AI tools can do, but how their use affects trust, decision-making and employee well-being. APA’s new guide gives leaders a practical framework for introducing AI thoughtfully, with people at the center of the process.”
The share of workers reporting pressure from their employers to use AI at work is rising. Yet, according to APA’s 2026 Work in America Survey released earlier this year, less than 40% of workers for whom this question applied reported knowing whether their employer had an official policy about acceptable AI use.
To address the adoption-policy mismatch, APA’s new guide offers a practical checklist with seven ways to help organizations introduce, use and evaluate AI tools responsibly. It also offers examples of what effective AI tool implementation can look like, as well as identifies the red flags to watch out for. Rather than providing universal instructions or telling organizations whether or not to use AI tools, the guide is designed to help organizations that choose to introduce these tools do so thoughtfully.
Developed with input from an advisory panel of experts in human-technology interaction and clinical, social, legal, cognitive and industrial-organizational psychology, the guide’s suggested practices include:
Matching safeguards to the stakes of potential errors and thinking through consequences and scenarios that may lie ahead;
Ensuring that AI is introduced through a fair process, including consulting with those who will be using the new system and those who will be affected by the outputs of that system; and
Protecting opportunities for people to develop deep expertise, share knowledge and learn from experience.
“Purchasing an AI tool and telling people to use it without further direction is not a sound strategy,” said Dennis P. Stolle, J.D., Ph.D., executive lead psychologist for applied psychology at APA. “Organizations need to invest in the people who will work with the technology, not just the technology itself. That means giving workers the information they need to judge where AI tools are reliable and where they are not, making sure human oversight is meaningful and rethinking how jobs are designed.”