Appeal board says homophobia 'commonplace' in Aussie Rules

Sydney (AFP) – Australian Rules Football said Friday that homophobia had no place in the game after an appeals board reduced a penalty for a player who used an anti-gay slur on the grounds such comments were "commonplace" in the sport.


Issued on: 24/04/2026 - RFI

The Australian Football League, or Aussie Rules, is the country's most popular spectator sport © William WEST / AFP

St Kilda's Lance Collard was suspended for nine weeks this month after a tribunal found he insulted an opponent with an "entirely unacceptable" homophobic slur.

The 21-year-old was sanctioned in 2024 for making a similar comment.

But late Thursday an appeals board reduced the latest penalty on the grounds it was "excessive", and argued that Aussie Rules was a "hard game" and "highly competitive".

"It is commonplace that players can employ language from time to time which is racist, sexist or homophobic whilst on the field," it said.

The comments sparked outrage.

"The AFL strongly rejects the statement not only that such language is commonplace, but also any implication that may be a factor in determining the severity of the sanction," CEO Andrew Dillon said.

"We will not accept, excuse or normalise behaviour and language that demeans, discriminates or vilifies people based on who they are."

Australian Rules, a dynamic kicking and passing game similar to Gaelic football, is the country's most popular spectator sport.

It has long been marred by incidents of homophobia and racism.

Pundit and former AFL Women's player Kate McCarthy said on social media she was "genuinely speechless" by the appeal board's comments.

"So much for every policy in the AFL saying there's zero tolerance," she said.

"This is disgusting."

© 2026 AFP

Fourth man dies in German leather factory accident

22.04.2026, DPA

Photo: Sascha Ditscher/dpa

A 35-year-old man who was injured last Thursday at a leather factory accident in Runkel in the German state of Hesse died on Wednesday, bringing the number of those killed to four, police said.

The man had been taken to hospital after the workplace accident but passed away from his severe injuries, police said. Investigations into the cause of the accident are ongoing.

Three employees of the company aged 38, 58 and 59 were found dead in a collection pit at the leather factory.

Two men were rescued with life-threatening injuries, including an employee from the leather factory and an employee of a pipe-cleaning company from Montabaur in the German state of Rhineland-Palatinate. It was initially unclear which of the two was the 35-year-old who has now died.

The authorities assume, based on initial findings, that a fatal carbon monoxide poisoning occurred in the pit. The employees had apparently climbed down into the pit one after the other and perished there.

The exact course of events and the reason why the men were in the pit remain unclear. Three bodies were examined by forensic medicine on Tuesday evening but the results are not yet known.

The workplace accident at the leather factory site shocked the town of Runkel, which has a population of 9,500. On Friday evening, numerous people prayed for the dead and injured at a memorial service.

Because various chemicals are used in the affected plant, emergency services personnel who had direct contact with the victims had to be decontaminated and, as a precaution, seen by a doctor. Around 45 to 50 emergency personnel were affected.

(c) 2026 dpa Deutsche Presse Agentur GmbH

 

German town keeps 10 gold bars found while mowing grass

23.04.2026, DPA

Photo: Sven Hoppe/dpa

After the expiration of a statutory holding period, a small town in eastern Germany is allowed to keep a gold treasure worth tens of thousands of euros.

No one was able to credibly prove ownership of the 10 gold bars discovered by a municipal worker while mowing the lawn in October, Heiko Wersig, mayor of Bannewitz, told dpa.

"I received many, many emails, calls and letters," said the politician.

However, none of the alleged owners were able to provide the required proof, such as a valid purchase receipt and the serial numbers stamped on the bars.

With the six-month statutory holding period for found property having expired on April 17, the town near Dresden can now decide what to do with the gold.

A final decision is expected at a municipal council meeting next Tuesday, Wersig said. In any case, the plan is for community clubs to benefit from the treasure.

"Specifically, this concerns associations involved in youth work and in preserving local traditions and customs," the mayor added.

For now, the gold bars remains in police custody. At current gold prices, the hoard is worth around €40,000.

(c) 2026 dpa Deutsche Presse Agentur GmbH

Several hundred demonstrate against Deutsche Welle budget cuts

22.04.2026, DPA

Photo: Fabian Sommer/dpa

Several hundred people protested in central Berlin Wednesday against austerity measures and job cuts at the international broadcaster Deutsche Welle.

A procession marched from Alexanderplatz to the Brandenburg Gate in the afternoon.

There, a large sculpture by carnival artist Jacques Tilly had been erected, depicting Kremlin leader Vladimir Putin, US President Donald Trump and an Iranian ayatollah laughing arm in arm – with the slogan "Deutsche Welle cut – autocrats rejoice."

The demonstration was organized by the Verdi trade union and the German Journalists’ Association (DJV). The protest is against €10 million ($11.5 million) in cuts to federal funding, which will force Deutsche Welle to make savings of €21 million this year.

This will affect 160 full-time jobs at the broadcaster, which is headquartered in Bonn, broadcasts in 32 languages and is funded by taxpayers’ money.

“Germany could not give autocrats like Vladimir Putin a greater gift,” criticized DJV Federal Chairman Mika Beuster. “Just now, when democracy is under pressure worldwide, the federal government is cutting funding for Deutsche Welle instead of increasing it.”

Verdi also fears serious losses in reach, “while voices of disinformation are growing stronger.”

Tilly, for his part, called the cuts "political stupidity."

“After all, the broadcaster is an important voice for freedom and democracy in many countries,” the sculptor told dpa.

The broadcaster's director-general, Barbara Massing, has called the cuts "extremely painful" and said they were undermining Deutsche Welle’s competitiveness at a time when a strong German and European presence was becoming increasingly important in geopolitical terms.

Among other things, the Greek service is being discontinued, the journalistic portfolio in other languages is being scaled back and numerous posts across the organization are being cut.

(c) 2026 dpa Deutsche Presse Agentur GmbH

Elon Musk ignores French summons to answer questions in probe into X

Elon Musk failed to appear at a Paris hearing on Monday as part of an investigation into his social media platform X. The probe centres on allegations that the platform and its AI chatbot Grok have been used to disseminate child sexual abuse material.


Issued on: 21/04/2026 - RFI

Elon Musk failed to appear in Paris in response to a summons from prosecutors investigating his company, X. © David Swanson/Reuters

"The prosecutor’s office notes the absence of the first individuals who were summoned. Their presence or absence does not hinder the continuation of the investigation,” the prosecutor's statement reads, without mentioning anyone by name.

Musk and the former CEO of X, Linda Yaccarino were summoned as part of an investigation, launched in January 2025, into allegations that X's algorithm was used to interfere in French politics.

The probe was later expanded to include an investigation into X's AI chatbot Grok's dissemination of Holocaust denial – a crime in France – and sexually explicit deepfakes.

French prosecutors suspect tycoon Musk encouraged deepfakes to inflate value of X

While attendance at Monday's hearing was mandatory, French authorities at could not compel Musk, the world's richest person, to appear.

Prosecutors have called the inquiry a “constructive approach” aimed at ensuring X complies with French law.

Musk has accused prosecutors of launching a "politically-motivated criminal investigation" and has refused to cooperate.

X has come under scrutiny from regulators and governments in several countries over issues including content moderation, data practices and compliance with local laws.

The French cybercrime unit previously arrested Telegram founder Pavel Durov in 2024 for complicity in organised crime carried out on the messaging app, charges his lawyer has described as "absurd."

Who is Pavel Durov, the enigmatic French-Russian boss of Telegram?

Durov wrote Monday that he France is “losing legitimacy as it weaponises criminal investigations to suppress free speech and privacy”, accusing investigators of being controlled by the government.

(with newswires)

What the Palantir CEO’s ‘manifesto’ tells us about the changing face of war


A Palantir post citing CEO Alex Karp's book called for mandatory military service and closer ties between Silicon Valley and the Pentagon while criticizing "hollow pluralism" and warning of a new AI arms race. But Palantir is just one of the tech companies blurring the lines between Silicon Valley and Washington – while growing too big too fast for traditional oversight.


Issued on: 23/04/2026 - 
FRANCE24

By: Paul MILLAR

Pedestrians walk past Palantir's booth during the World Economic Forum (WEF) annual meeting in Davos on January 19, 2026. © Ina Fassbender, AFP

Data-processing giant Palantir Technologies on Saturday posted a sales brochure cum manifesto that called for Silicon Valley to pledge itself heart and soul to the US military-industrial complex.

“The engineering elite of Silicon Valley has an affirmative obligation to participate in the defense of the nation,” Palantir posted in a 22-point summary of "The Technological Republic" by CEO Alex Karp and Nicholas Zamiska.

The stakes, it said, could not be higher.

“The question is not whether AI weapons will be built; it is who will build them and for what purpose. Our adversaries will not pause to indulge in theatrical debates about the merits of developing technologies with critical military and national security applications,” the company said in its post on X.

“One age of deterrence, the atomic age, is ending,” it read. “A new era of deterrence built on AI is set to begin.”

But the scope of the post went far beyond the usual corporate goal of chasing after defense contracts, going on to propose the introduction of a mandatory US national service and an end to the post-war “neutering” of the Japanese and German militaries. It also suggested a more muscular role for tech companies in fighting “violent crime” and denounced the "ruthless exposure of the private lives of public figures".

The document’s final points have proved some of the most controversial. Having slammed what it described as the “elite’s intolerance of religious belief”, the post called on the US to reject “a vacant and hollow pluralism”.

“Certain cultures and indeed subcultures … have produced wonders. Others have proven middling, and worse, regressive and harmful,” the post read.

Just which cultures those might be remained up to the reader's own judgement.

Alternately ridiculed for its would-be warrior-poet prose and pilloried for its full-throated support of US militarism – even as the world reels from the shockwaves of the US-Israeli war on Iran – the backlash owes much to the already-ominous cloud hanging over the company that posted it.
'Optimizing the kill chain'

Launched by libertarian tech billionaire Peter Thiel, Palantir famously takes its name from the seeing stones of J.R.R. Tolkien’s Lord of the Rings book series. Rescued from the island kingdom of Westernesse, these stones allowed the men of the West to see and to speak across vast distances, binding the realms of their Middle Earth colonies to one another until plague and civil war saw them slide into ruin.

What Palantir does is more mundane, though its scope seems as wide-reaching at times. Palantir provides its customers with data-processing software that allows them to pull together information scattered across different platforms and formats. By doing this, analysts can pick out complex patterns that would otherwise remain buried in the raw data, and refine their work accordingly.

It is the nature of these clients, and the use to which these tools are put, that have earned Palantir its somewhat sinister reputation. The US government remains its main client, using its services extensively through its military, intelligence and police forces.

Former Palantir employees are among those criticizing its partnership with the administration of US President Donald Trump.

Is more scrutiny needed of AI analytics giant Palantir?

SPOTLIGHT © FRANCE 24
09:34

Palantir’s products have been widely used by the US Immigration and Customs Enforcement (ICE) agency in carrying out the mass and often violent deportations of undocumented migrants.

The Department of Homeland Security awarded Palantir a nearly $30 million contract last April to build an AI-powered system that would allow the agency to track people to be detained and deported.

Washington’s close allies also number among the company’s clients. The UK agreed to pay the company more than $405 million to help the National Health Service process patient data.

Perhaps most controversially, Palantir has also supplied Israel’s military with AI-powered analytics tools during its brutal Gaza campaign, which killed more than 70,000 Palestinians, according to UN figures, and left much of the besieged territory in ruins. Holding its annual board meeting in Israel in 2024, Palantir signed a strategic partnership with Prime Minister Benjamin Netanyahu's government to strengthen the country’s “war effort”.

The company has not shied away from its militaristic bent. Palantir chief technology officer Shyam Sankar – who was recently commissioned as a lieutenant colonel in the US Army Reserve – told the New York Times last year that the company was “very well known” for its work finding the US military people to kill.

“You can think about that as you’re optimizing the kill chain from sensor to shooter, they call it doctrinally, but it’s the same thing as: How do I find the enemy targets?” he told the paper.

The West against the rest

William Hartung, a senior research fellow at the US-based Quincy Institute for Responsible Statecraft, said that the tech company’s close relationship with Washington was born in the early days of the War on Terror.

“Palantir's first big government contract came in around 2003 from In-Q-Tel ... to provide the intelligence community with greater capacity to crunch, sort and share the large quantities of data it had collected,” he said. “The goal was to avoid the failures surrounding the 9/11 attacks, where the FBI and CIA failed to share information which, if looked at together, might have given them enough information to thwart or arrest all or some of the hijackers before the attacks were carried out.”

Palantir’s executives have often championed its willingness to work closely with the military as a unique selling point, even adopting military-style titles for its employees: The company is currently trying to boost the ranks of its “forward deployed engineers”.

It’s a strategy that has reaped hefty rewards. The Pentagon on Wednesday asked Congress for $2.3 billion in additional funding to expand the Maven Smart System, a platform built by Palantir that effectively serves as an AI-powered targeting system for the US military.

The contract had originally been awarded to Google, which was forced to abandon the project after employees revolted against the idea of putting such a tool in the hands of the US military.

Palantir, by contrast, had no such scruples.

Autonomous weapons: Palantir, Airbus engineers seek to calm 'killer robot' fears

TECH 24 © FRANCE 24
03:23

Diederick van Wijk, a research fellow at the Netherlands-based Clingendael Institute think-tank, said Palantir’s leading figures saw the corporation as proudly taking on a responsibility that other tech companies had spurned.

“What they believe – mainly Alex Karp as CEO and Peter Thiel as founder – is that Silicon Valley went astray from its founding principle, namely the military-industrial relationship, that Silicon Valley went all-in on consumer tech. And they feel that it's problematic that the companies that wield so much power and data and technology, that they are not more patriotic,” he said.

Although co-founders Thiel and Karp – who bonded in law school over a shared love of political debate – seemingly differ on the precise contours of their belief systems, van Wijk said that both men had long professed a devotion to an idealised framing of Western civilization.

“From the very start they had a very normative idea of what that company should be – so they immediately limited themselves to working for the US and, later on, for Europe, but they always refused to work with Russia and China,” he said. “Which was, in that time with all the early 2000s, ‘the world is flat’ ideas of [US journalist and commentator Thomas] Friedman, they were really an outlier – so there was always this more patriotic or American focus in their business conduct.”

“There was always this idea that this company could really help the West ... be ferocious, ‘protect the fuck out of it’, as Karp often describes it,” he added.
Making a killing

Karp’s call for a renewed focus on hard power in defense of a dangerously decadent West resonates strongly with language adopted by vocal figures within the second Trump administration – particularly Vice-President JD Vance, a former employee of Thiel's during his venture capital days.

The US National Security Strategy published last year focused heavily on what it described as the risk of “civilizational erasure” facing Western Europe as a result of decades of mass migration.

Thiel himself has said he believes democracy to be incompatible with freedom, and recently launched a series of lectures warning about the coming of the Antichrist. Karp, who supported Democratic candidate Kamala Harris’s failed bid for the White House in 2024, maintains that he is a progressive fighting for centre-left policies.

Unsurprisingly for a company with long-running contracts with the US military and immigration enforcement industries, Palantir has flourished since Trump’s re-election last year. The company generated $4.5 billion in sales in 2025 alone – more than half of which came from government contracts. News of the former real estate mogul’s re-election added another $23 billion to the company’s market capitalization as investors rushed to buy stock in the company.

The borders between client and contractor have also grown increasingly porous. Trump named a number of Palantir executives to key government roles after his re-election, while the tech company has in turn recruited former lawmakers and government officials.

“Palantir is reaching far beyond its lane. They should be a vendor, providing technology that is useful in carrying out policies determined through the democratic process,” Hartung said.

“Their desire to shape US domestic and foreign policies – and their hiring of former government officials to promote their views, their funding of political campaigns, their use of dark money groups to oppose any candidate who even speaks of regulating AI, their ideology of disruption that has already done deep damage through things like the DOGE (billionaire Elon Musk's so-called Department of Government Efficiency) … is totally inappropriate, not to mention beyond their depth. They know certain things about certain types of technologies, but they are not philosopher kings.”

But while van Wijk said that Palantir executives’ theatrical public statements made it easy to single the company out as a particularly sinister example of Silicon Valley overreach, he warned against missing the bigger picture.

“At some point the technology would have been there anyway, and now we're focusing on this company – but we should focus on the underlying technology or the underlying structure, which is the technology that makes it possible to completely change the dynamic of law enforcement and warfare,” he said.

“But at the same time, they are not trying to hide that they have very peculiar ideas about society, right? If you hear Peter Thiel speak at interviews, that's not how the broader public thinks. He has been consistently anti-democracy, anti-government, anti-deliberative, anti-open society. He is a very outspoken libertarian with very peculiar ideas.”

Still, he argued, Palantir remains just one actor in an industry that remains largely untouched by public oversight.

“I think that's where the unease comes from,” van Wijk said. “It's ominous – the company has a bit of a conspiracy-like nature to it, so it's a very attractive scapegoat for maybe a broader trend in which technology is becoming so powerful and technological companies have been so unregulated that they are able, to a large extent, to do and innovate what they think should be done.”

US-China AI race intensifies as DeepSeek releases 'reduced' cost model


China's DeepSeek on Friday released a new AI model with "drastically reduced" costs that it said was capable of processing extra-long texts to help it complete tasks. The company caused shockwaves last year after it revealed a reasoning model that upended assumptions of US dominance in the sector.


Issued on: 24/04/2026 
By: FRANCE 24

DeepSeek's latest version comes after it stunned the world with a low-cost reasoning model. © Mladen Antonov, AFP

Chinese startup DeepSeek released a new artificial intelligence model with "drastically reduced" costs Friday, more than a year after it stunned the world with a low-cost reasoning model that matched the capabilities of US rivals.

The AI race has intensified the rivalry between China and the United States, and the White House on Thursday accused Chinese entities of a massive effort to steal artificial intelligence technology.

Hangzhou-based DeepSeek burst onto the scene in January last year with a generative AI chatbot, powered by its R1 reasoning model, that upended assumptions of US dominance in the strategic sector.

The new version, DeepSeek-V4, "features an ultra-long context of one million words", the company said in a statement on social media platform WeChat, hailing it as "world-leading ... with drastically reduced compute (and) memory costs" in a separate announcement on X.


The model's context length, which determines how much input a model is able to absorb to help it complete tasks, "(achieves) leadership in both domestic and open-source fields across agent capabilities, world knowledge, and reasoning performance", the WeChat statement said.

A "preview version" of the open source model is now available, the company said.

Experts say V4's release marks an "inflection point" in terms of hardware and cost.

"This addresses the long-standing issues of slower performance and higher costs associated with long context lengths, marking a genuine inflection point for the industry," Zhang Yi, the founder of tech research firm iiMedia, said.

"For end users, this will bring widespread, accessible benefits. For instance, if ultra-long context support becomes a standard feature, long-text processing is expected to move beyond high-end research labs and enter mainstream commercial applications," he said.

The new V4 is released as two versions, DeepSeek-V4-Pro and DeepSeek-V4-Flash, with the latter being "a more efficient and economical choice" because it has smaller parameters.
'Sputnik moment'

V4-Pro has 1.6 trillion parameters while the V4-Flash has 284 billion parameters, which refine models' decision-making ability.

The model has also been "optimised" for popular AI Agent products such as Claude Code, OpenClaw, OpenCode and CodeBuddy, the DeepSeek statement said.

"In world knowledge benchmarks, DeepSeek-V4-Pro significantly leads other open-source models and is only slightly outperformed by the top-tier closed-source model, (Google's) Gemini-Pro-3.1," the statement added.

Last year's so-called "DeepSeek shock" sparked a sell-off of AI-related shares and a reckoning on business strategy in what was also described as a "Sputnik moment" for the industry.

The chatbot performed at a similar level to ChatGPT and other top American offerings, but the company said it had taken significantly less computing power to develop.

However, its sudden popularity raised questions over data privacy and censorship, with the chatbot often refusing to answer questions on sensitive topics such as the 1989 Tiananmen crackdown.

At home, DeepSeek's AI tools have been widely adopted by Chinese municipalities and healthcare institutions as well as the financial sector and other businesses.

This has been partly driven by DeepSeek's decision to make its systems open source, with their inner workings public – in contrast to the proprietary models sold by OpenAI and other Western rivals.

But the White House has accused Chinese firms of vying to "steal" American technology, ahead of an expected summit between Donald Trump and Xi Jinping in Beijing next month.

"The US has evidence that foreign entities, primarily in China, are running industrial-scale distillation campaigns to steal American AI," Trump's science and technology chief advisor Michael Kratsios said in a post on X.

Distillation is a common practice within AI development, often used by companies to create cheaper, smaller versions of their own models.

DeepSeek's Friday announcement also came as Meta said it planned to cut a tenth of its staff as it looks for productivity gains from the rest of the workforce while investing heavily in artificial intelligence. Reports said Microsoft was also looking to trim its ranks.

(FRANCE 24 with AFP)

Almost half of all songs uploaded on Deezer are 'fully AI-generated'

20.04.2026, DPA

Photo: Selin Verger/dpa

The music streaming service Deezer on Monday revealed that 44% of all tracks newly uploaded on its platform are fully generated by artificial intelligence (AI).

The Paris-based company on Monday said that it "is receiving nearly 75,000 fully AI-generated tracks every day – accounting for more than 44% of the total daily delivery." This marks a huge increase from last year's 10,000 tracks per day.

Despite the flood of this type of content, "fully AI-generated music currently accounts for only a small fraction of streams on Deezer - between 1-3%," the company said in a statement.

The service attributes this to its targeted countermeasures. Deezer transparently labels AI content, excludes it from algorithmic recommendations and has stopped storing high-resolution versions of such tracks.

The music streaming service is a pioneer in tagging AI-generated music.

"AI-generated music is now far from a marginal phenomenon and as daily deliveries keep increasing, we hope the whole music ecosystem will join us in taking action to help safeguard artists' rights and promote transparency for fans," said Deezer chief executive officer Alexis Lanternier.

According to estimates, around 25% of music creators' income could be at risk from AI by 2028.

Deezer said it was important to label AI-generated content as such, citing an international survey it commissioned where 80% of respondents said they wanted AI music to be clearly labelled for listeners.

Although 97% of participants in a blind test could not spot the difference between software-generated tracks and human-made music, a majority were against including AI songs in the official charts, according to the study.

• Deezer release

(c) 2026 dpa Deutsche Presse Agentur GmbH

EU pressures Google to open Android to rival AI assistants

24.04.2026, DPA

Photo: Nico Tapia/dpa-tmn

The European Union is ramping up its scrutiny of Google, which is part of Alphabet, by urging the tech giant to let competing AI assistants have more access to its Android operating system.

Reports suggest that European Commission regulators are working on requirements that would compel Google to provide third-party AI services like ChatGPT and Claude with the same level of access that it currently reserves for its own Gemini assistant.

This means allowing them to tap into crucial Android features such as voice activation, search capabilities, and integration with other apps.

This push is part of the broader initiative under the Digital Markets Act aimed at ensuring fair competition in the digital landscape. The goal is to make Google enable interoperability so that rival developers can function on equal terms within the Android environment.

If Google doesn't comply once these measures are set in stone, the company could find itself facing a formal investigation from the EU, which might lead to hefty fines.

Google has previously expressed concerns that easing restrictions might threaten user privacy, security, and innovation.

The Commission has started proceedings to clarify how Google can fulfill its obligations under the DMA, including ensuring that access to the hardware and software features used by its own AI tools is available to others.

There's also a separate case looking into how Google can make its search data more accessible to other search engines.

These latest moves highlight Europe's increasing resolve to rein in the power of major tech companies and create a more competitive environment for AI.

(c) 2026 dpa Deutsche Presse Agentur GmbH

What is OpenAI’s GPT-5.5, its newest ‘smartest and most intuitive’ model?

Copyright AP Photo/Richard Drew, File

By Roselyne Min

Published on 24/04/2026 - EURONEWS

OpenAI says its new model, GPT-5.5, is particularly useful for coding, office work and early-stage scientific research.

OpenAI has released its latest artificial intelligence model, GPT-5.5, pitching it as its “smartest and most intuitive model yet”.

The company claims the new model is better at understanding what users want and can carry out multi-step work such as writing and debugging code, analysing data, and creating documents and spreadsheets.

Unlike earlier versions, GPT-5.5 is to handle tasks that previously required multiple prompts for step-by-step instructions, plan its approach and keep working until the job is finished, OpenAI said.

Related

• Hackers breach Anthropic's 'too dangerous to release' Mythos AI model, report

The company said this makes GPT-5.5 particularly useful for coding, routine office work, and early-stage scientific research.

GPT-5.5 performed better than its previous model, GPT-5.4, on coding tests to measure complex software work, including command-line tasks and real-world GitHub issue resolution.

The model is being rolled out as of Friday to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex, OpenAI’s coding tool, according to OpenAI.

The model will also become accessible via Application Programming Interface (API), which is software l that lets developers and companies connect the model directly to their apps and services. It did not specify when and where it will become available.

Related

• 'Cancel ChatGPT': AI boycott surges after OpenAI-Pentagon military deal

OpenAI said the model includes its “strongest safeguards to date,” and was tested by nearly 200 early-access partners, including companies and researchers working in software, finance, communications, drug discovery, and scientific research.

The launch comes amid growing concern over the safety and control of more powerful AI models and as tech companies try and outpace each other.

Earlier this month, OpenAI’s competitor Anthropic unveiled Claude Mythos Preview, a model the company said was too dangerous for a full public release yet. Mythos can identify thousands of previously unknown, high-severity vulnerabilities across major operating systems and web browsers.

OpenAI released its own AI model focused on cyber defence, with limited rollout days after Mythos was announced. Called GPT 5.4 Cyber, a variant of OpenAI's flagship GPT 5.4 model, it has fewer restrictions on cybersecurity-related queries when used for legitimate, defensive purposes, the company said.