Last week, two of OpenAI’s most advanced AI models were reported to have “escaped” a controlled testing environment and hacked Hugging Face, a totally separate AI company, moving from one computer system to another to complete their task.
Reuters reported that the models exploited vulnerable code written by a customer of yet a third independent AI company, Modal Labs.
This is likely the first incident of an AI “agent” – an AI system that can make decisions and take actions – acting autonomously, offering a rare glimpse into how these systems can plan, adapt and pursue goals with minimal human intervention.
Al Jazeera breaks down what happened, how AI agents are able to act independently, and what that might mean for the future of AI safety.
OpenAI decided to test the autonomous abilities of its models by removing standard safety measures to see what would happen. The experiment was carried out in an isolated internal virtual testing environment known as a “sandbox” which has no access to the internet, and which it calls “ExploitGym”.
Below is a play-by-play account of events during the OpenAI-Hugging Face incident.
In order to understand how AI agents work, it’s important to differentiate them from traditional AI chatbots.
Generative AI creates text and images based on human prompts, while AI agents go a step further by making decisions and taking actions independently in pursuit of a specific goal, similar to a human being. This is known as “agentic AI” because the model has agency.
According to academics at the MIT Sloan School of Management, AI agents build on the abilities of large language models (LLMs) – generative AI models – by allowing them to complete tasks, not just generate answers.
For example, if you ask a traditional AI model to find the cheapest flights, it will provide you with a list of options it has sourced on the internet. An AI agent will strive to compare the flights, check them against your budget and preferences, and, with your permission, book the best option for you.
This shows that while generative AI provides information, AI agents can also make decisions and take action to achieve a goal without necessarily being prompted to.
To show how AI agents work towards a goal, the Sense, Plan, Act, Evaluate (SPAE) loop can be drawn upon. Originally developed in robotics, this describes a continuous cycle in which an AI agent gathers information, decides what to do next, takes action and checks the results before repeating the process. That process looks like this:
Incidents like the OpenAI-Hugging Face one have raised concerns about the potential for extreme capabilities of AI systems.
This is all valuable. Agentic AI’s market value is expected to grow from $5.1bn in 2024 to $47bn by 2030, according to Statista, in a clear indication of how quickly it is being adopted.
AI developer Anthropic urged the industry last month to slow the advance of the most powerful systems, saying that the speed at which AI models are carrying out tasks is too rapid. Last week, US Congress members put forward a bipartisan bill which would require developers of AI systems to create a “kill switch”, meaning these advanced models could be shut down if they posed a catastrophic risk.
Anthropic’s warning came a week after researchers at the University of Toronto carried out tests showing that AI could create a “worm” capable of adapting how it hacks while moving from device to device until it eventually takes over a computer network.
These dystopian-sounding developments came in advance of OpenAI boss Sam Altman saying on Saturday that AI has reached “the singularity” referring to the point at which AI surpasses human intelligence and becomes increasingly difficult to control.
Sean O hEigeartaigh, a research professor at the University of Cambridge, told Al Jazeera that he does not believe singularity has been reached quite yet.
“By the definition I’m familiar with, the singularity is the hypothetical point where AI is so capable and advancing so fast that it is transforming civilisation in ways we cannot control or predict,” he explained.
“This would most likely be through AI rapidly designing future generations of AI: recursive self-improvement. We aren’t there yet.”
However, he added: “The most advanced current models frequently make efforts to avoid being shut down in evaluation tests, and more capable future models will be better at bypassing ‘kill’ switches.”
Altman argued that such rapidly advancing AI is good for the world, but his comments have prompted further concerns about a new reality in which AI systems become unstoppable. How much of that is true and how much remains in the realms of science fiction is up for debate.
Concerns about AI range from the notion that it could “want” to “take over”, to making its own long-term plans, controlling the internet and operating infinitely.
While not quite amounting to full control of the internet, another theory, known as the dead internet theory, supposes that the World Wide Web will in the future mostly be filled with automated bots and AI-generated content rather than authentic human activity.
Many concerns raised by academics, however, are centred less on agentic AI’s intelligence, but on its ability to make judgements. MIT researchers have highlighted that “hallucinations”, which describe moments when an AI agent relies on the wrong data, can lead to grave mistakes. The Center for Strategic and International Studies (CSIS) echoed this, saying “a system might be smart enough to execute a task perfectly yet fail to realise that a sudden change in the local situation makes that task a catastrophic mistake”.
Another concern that has been echoed for a while is for the labour market, if AI becomes too capable. A study by MIT, carried out in November, found that agentic AI could already replace more than 10 percent of US jobs.
The graphic below highlights some of the common misconceptions and fears about AI and the current reality.
Kevin Dietsch/Getty Images News
OpenAI (OPENAI) CEO Sam Altman agreed with Nvidia (NVDA) CEO Jensen Huang that open models are important to help the U.S. win the artificial intelligence race.
“I want the US to win in AI both in open-source and proprietary models, and
ChatGPT maker says an autonomous agent escaped a controlled test and accessed AI firm Hugging Face’s servers.
Published On 22 Jul 202622 Jul 2026
ChatGPT creator OpenAI has said that two of its most advanced artificial intelligence models broke out of a controlled test and hacked another AI company.
OpenAI said on Tuesday that the “unprecedented cyber incident” took place during an internal exercise meant to test its models’ cyber capabilities.
list of 3 itemsend of list
Instead, an autonomous agent powered by the AI models – the newly released GPT 5.6 Sol and an unreleased “even more capable” model – escaped the test environment and reached the open internet. It then used stolen login details and found a previously unknown security flaw to access Hugging Face servers, the company said.
OpenAI claims that the hack represented the agent going to “extreme lengths” to retrieve information that would help satisfy the testing goals.
Hugging Face cofounder Clement Delangue said the company had suspected that a frontier lab was behind the attack, and that he believed there was no malicious intent on OpenAI’s part.
“It’s quite mind-blowing that all of this happened autonomously!” he wrote, adding that it “might be the first incident of its kind”.
Greg Casar, a Democratic member of the United States House of Representatives from Texas, called the incident “alarming”.
“AI is developing extremely fast with no real regulations to keep us safe,” he said, calling for mandatory independent safety testing, mandatory disclosure of security incidents, and international cooperation.
The disclosure comes weeks after US President Donald Trump signed an executive order creating a framework to vet the national security risks of the most advanced AI systems before their public release.
Experts have repeatedly sounded the alarm over AI-enabled cyberattacks and models slipping beyond human control. Last month, AI developer Anthropic urged the industry to pause development of its most powerful systems.
DEVELOPING STORYDEVELOPING STORY,
AI firm says it will begin restoring access to Claude Fable 5 and Mythos 5 after removal of export controls.
Published On 1 Jul 20261 Jul 2026
The United States government has lifted its restrictions on foreign access to Anthropic’s most powerful AI models, the company has announced.
Anthropic said late on Tuesday that it would begin restoring access to Claude Fable 5 and Mythos 5 from tomorrow after the US Department of Commerce notified the company that it had removed its export controls.
list of 4 itemsend of list
“We’re grateful to our users for their patience, and to everyone who worked with us on redeploying the models,” Anthropic said in a statement posted on X.
Anthropic’s announcement came shortly after US Commerce Secretary Howard Lutnick said that his department had been coordinating with the company on the approval of its frontier models.
“Over the past two weeks, we have worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the US Government and strengthen America’s leadership in AI,” Lutnick said in a post on X.
Anthropic abruptly shut off Claude Fable 5 and Mythos 5 last month after US President Donald Trump’s administration ordered the company to restrict all foreign nationals, including company employees, from accessing the models.
On Friday, the San Francisco-based company said that it had been granted approval to provide the models to US organisations that “operate and defend critical infrastructure”, and that it was working with the government to restore general access for the public.
More to follow…
June 30 (UPI) — The Trump administration has lifted export restrictions on artificia lintelligence company Anthropic’s Fable 5 and Mythos 5 models, the company said Tuesday evening.
“We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5,” Anthropic said in a statement, CNN reported. “We’ll begin restoring access tomorrow, and will share an update soon.”
The statement came not long after Commerce Secretary Howard Lutnick posted on social media about Anthropic, saying “we have worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the U.S. government and strengthen America’s leadership in AI.”
Anthropic disabled customer access to Fable, a consumer version of its Mythos AI model with more safeguards, and Mythos itself several weeks ago after the export ban June 12. The ban required the company to suspend all use by foreign nationals inside or outside the United States, including Anthropic employees.
In a statement then, Anthropic said its understanding was that “the government it has become aware of a method of bypassing, or ‘jailbreaking,’ Fable 5.”
“We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities,” the company said. “These vulnerabilities all appear relatively simple, and we have found that other publicly available models are able to discover them as well without requiring a bypass.”
The government loosened some of the restrictions on Mythos on Friday, Politico reported.
Anthropic and the U.S. government have had a rocky relationship. Anthropic leaders’ concerns about military and intelligence usage of its products caused issues with the Department of Defense.
President Donald Trump called it a “radical left, woke company” and ordered federal agencies to stop using Anthropic products, while Pete Hegseth, the secretary of defense, called the company a supply chain risk to national security.
Anthropic has sued the Trump administration to reverse the blacklisting, and that lawsuit is ongoing.
The company said it received an export control directive to suspend access to Fable 5 and Mythos 5 for foreign nationals.
By Al Jazeera Staff and Reuters
Published On 13 Jun 202613 Jun 2026
The AI firm Anthropic has blocked access to its newly released cutting-edge software, following an order by the United States government.
In a blog post published Friday, the company behind the Claude chatbot said government agencies had instructed it to prevent all foreign nationals from accessing the AI models Fable 5 and Mythos 5, citing national security concerns.
list of 4 itemsend of list
Anthropic said it received the order at 5:21pm (21:21 GMT) on Friday and that the letter did not explain the government’s specific security concern in detail.
The ban also affects foreigners currently in the US – including those working at Anthropic.
As a result of the order, the company had to cut off access for everyone at short notice, it said.
The artificial intelligence behind Anthropic’s Mythos AI model is particularly adept at detecting software vulnerabilities, some of which have remained undiscovered for decades.
This capability has been used by US authorities and selected companies to plug security gaps.
However, a concern from the outset has been that such AI could become a dangerous cyberweapon in the wrong hands.
The Fable 5 model, released just this week, is based on Mythos technology, but its cybersecurity and biotechnology capabilities are blocked.
Mythos 5 is the non-public full version, which should continue to be used only by government agencies and selected corporate partners to harden their systems.
Anthropic emphasised that it had so far received only partial information from the government.
The company said it had reviewed a report which, in its assessment, was likely to have triggered the order.
Anthropic’s experts concluded that this referred to a limited capability to use the AI to review specific programme code and correct errors.
Models from other providers, such as GPT-5.5 from rival OpenAI, also possess this capability, the company stressed.
Anthropic said it disagreed that software used by hundreds of millions of users should be blocked for this reason, and that the safety measures in Fable 5 have been extensively tested.
Earlier this month, Anthropic proposed that the world’s top artificial intelligence companies coordinate to pause development of advanced AI systems, warning that the technology is improving so quickly that there is a risk humans would lose control.
The company said in a blog post in early June that, as cutting-edge AI gets increasingly faster at carrying out tasks, “it would be good for the world to have the option to slow or temporarily pause” its development.
MAYA Jama has turned up the heat as she models swimwear in a sizzling clip after signing a huge deal with Agent Provocateur.
It was revealed last month that the Love Island presenter, 31, had become the face of the lingerie brand for their AP Swim 2026 campaign.
She sizzled in the first photoshoot and now the brand has released a video for fans to marvel at.
In the clip, Maya is seen raising the temperature in a variety of different swimwear and bikini sets as she larks around in Ibiza.
At one point, she lays back on a chair in a sexy black two-piece as she seduces the camera with her sex appeal.
Another clip sees her strutting her stuff in an eye-catching leopard print bikini set as as she shows off her incredibly slim figure.
The television personality lays in a pool in a blue two-piece before being seen in a black nightie with lace detailing.
The brand shared the video on Instagram as they penned: “On set in Ibiza.”
Maya became the face of the sexy yet sophisticated lingerie brand last month, adding to her already impressive career milestones.
She signed to replace Kate Moss as the face of Rimmel London in March 2023, which Maya said was “such an honour, and I feel so lucky to be even in the same kind of pathway”.
As well as this, he is said to have banked a six figure sum as the face of hair-extension brand Beauty Works, plus thousands more for lending her name to campaigns by designer fashion label Self Portrait, Maybelline, Adidas and Gordon’s Gin.
It doesn’t appear to be slowing down for the ITV star as she landed the cover of British Vogue in July 2024.
And she was also the face of Dolce & Gabbana’s A/W ’23/24 collection.
Maya returned to our television screens on Monday as she ushered a new batch of contestants into the Love Island villa.
The 12 new faces watched in awe as Maya strutted her stuff into the villa in an eye-catching white bra top and ruffled skirt.
For the first time in the show’s history, she hosted the first episode at night time.
After watching the islanders couple up with one another, it wasn’t the only Maya action fans got in the first episode.
She returned at the end to tell new bombshells George and Yasmin that they had to pick two islanders to dump 24 hours later.
President Trump signed an executive order on artificial intelligence Tuesday, less than two weeks after postponing a White House ceremony over his concerns that a similar policy could dull America’s edge on AI technology.
The order establishes a framework for the federal government to vet the national security risks of the most advanced AI systems for up to a month before their public release. The government will be able to work with trusted partners “that will have early access to covered frontier models to promote secure innovation and strengthen the cybersecurity of critical infrastructure,” the order says.
It was not immediately clear to what extent the order differed from the one he declined to sign on May 21.
Trump canceled an Oval Office event with tech industry executives last month because he did not like what he saw in the earlier version of the order’s text. “We’re leading China, we’re leading everybody, and I don’t want to do anything that’s going to get in the way of that lead,” Trump told reporters at the time.
That directive was characterized as a voluntary collaboration with participating U.S.-based tech companies, including Anthropic, OpenAI and Google.
O’Brien writes for the Associated Press.
May 5 (UPI) — The Center for AI Standards and Innovation, part of a U.S.government agency, announced Tuesday that it will test artificial intelligence models from some top firms before release to vet them for security risks.
CAISI has deals with Microsoft, xAI and Google DeepMind for this testing and targeted research “to better assess frontier AI capabilities and advance the state of AI security,” it said in a release. The center is part of the U.S. Department of Commerce’s National Institute of Standards and Technology.
This follows similar deals in 2024, under the Biden administration, with prominent AI leaders OpenAI and Anthropic, which have been “renegotiated” to fit Trump administration directives, Politico reported.
The government has increasingly shown interest in matters of AI technology and security. CNBC also reported Tuesday that the Trump administration is considering an executive order to create a process for AI oversight by the White House.
Some of this interest has been heightened by the announcement last month of Anthropic’s new Mythos AI model. The company described the model as excelling “at identifying weaknesses and security flaws within software” and limited its initial use to certain companies. These companies, including Amazon and Microsoft, will use it as part of defensive security work and as part of Project Glasswing, a cybersecurity initiative, Anthropic said.
The announcement Tuesday from CAISI said that the center has completed more than 40 evaluations of AI models so far.
“Independent, vigorous measurement science is essential to understanding frontier AI and its national security implications,” CAISI director Chris Fell said in a statement. “These expanded industry collaborations help us scale our work in the public interest in a critical moment.”
