- Published
OpenAI has revealed a cyber-attack carried out by rogue ChatGPT agents went further than just one company.
asserted
attack → publish → company
Hugging Face was thought to be the only victim of the unprecedented hack - but OpenAI now admits its bot attacked several "publicly-available services".
asserted
bot → think → services
The out-of-control AI found four logins online which allowed it to access four separate, unnamed services.
asserted
it → find → services
Meanwhile, in an emergency briefing with hundreds of cyber security professionals, Hugging Face has described what it was like to be on the receiving end of the world's first fully autonomous AI hack.
asserted
it → describe → hack
The firm described how the AI worked at superhuman speed but also made strange decisions and mistakes that no human hacker would have made.
Hugging Face, which is like an app store for AI tools, said the hacking agents worked relentlessly with thousands of different methods trialled simultaneously.
The company first revealed that it had been hacked, external by someone using powerful autonomous AI on 16 July and reported it to police.
asserted
it → describe → police
Nearly a week later, OpenAI admitted it was its AI that had escaped a closed environment and attacked Hugging Face on its own during a test.
asserted
that → admit → test
It was trying to find the answers to a hacking exam it had been set by OpenAI, and targeted Hugging Face.
asserted
it → try → Face
On Wednesday OpenAI updated its statement to include the extra detail that the hack went further than first thought.
asserted
hack → update → detail
"The models identified and used publicly exposed credentials at the account-level on other publicly-available services.
asserted
models → identify → services
This includes four accounts on four services," the company said.
asserted
company → include → services
OpenAI did not clarify whether "publicly-available services" means companies - but it said the new attacks were not the same level of severity as the Hugging Face hack.
asserted
attacks → clarify → hack
On Tuesday, the industry body the Cloud Security Alliance (CSA) wrote-up a report , externalbased on the emergency meeting with Hugging Face on Friday - which Hugging Face itself has reviewed.
asserted
Face → write → which
"The agents followed inefficient routes and exhibited clumsy behaviours that no human would choose", the CSA wrote.
asserted
CSA → follow → that
The agents repeated actions that they had already completed - a sign of an agentic AI losing its thread and context.
asserted
AI → repeat → thread
The agents also hallucinated reams of incoherent commands and text and were sloppy and did not cover their tracks well.
asserted
agents → hallucinate → tracks
But among the errors and strange behaviour, Hugging Face warned the AI agents made brilliant technical moves and were able to rapidly adapt to new scenarios in the days-long hack.
asserted
agents → warn → hack
It took three days for them to be discovered inside the Hugging Face IT network and it took the company's AI and cyber-security experts many hours to contain and eject the AI agents - something standard companies might struggle with.
uncertain
companies → take → something
The company would not say how much the hack cost it but said staff worked for many hours to rebuild about a third of their infrastructure.
asserted
staff → say → infrastructure
Hugging Face has been praised for its transparency in telling the AI and cyber industry what happened.
asserted
what → praise → industry
The CSA warned the incident shows that AI "agents... find a way" - a reference to the film Jurassic Park, where dinosaurs escape their enclosures.
asserted
dinosaurs → warn → enclosures
"They are objective-driven, set their own sub-goals, adapt in real time to bypass defences, and operate with a machine-speed persistence that can overwhelm manual operations," the paper reads.
asserted
paper → drive → operations
Cyber security officer Ritesh Patel was on the Hugging Face briefing call with around 450 others and says the industry is working hard to address the new threat of rogue AI agents.
"This is the reality of autonomous agents powered by frontier models: they are relentlessly persistent, sometimes highly noisy, and will try every possible path to achieve their goal, which can easily overwhelm traditional defences," he said.
asserted
he → say → defences
Ethical hacker Valentina Palmiotti - better known as Chompie - reviewed the CSA report and says the way the agents hack might seem haphazard but it is clearly effective.
uncertain
it → know → report
"They throw out a bunch of stuff and see what sticks," she said.
asserted
she → throw → stuff
"But they also don't get bored, they don't sleep and can be infinitely tenacious.
asserted
they → get → ?
This is not the first time AI agents have been shown to go "rogue".
asserted
agents → show → ?
In the CSA's report it references previous examples like in September 2024 when an earlier model of ChatGPT escaped its container to get an answer it needed for another test.
asserted
it → reference → test
That event was contained in OpenAI's own IT systems and "largely celebrated at the time", the CSA noted.
asserted
CSA → contain → time
But "rogue" behaviour "is the standard, not the exception," the paper claimed.
asserted
paper → claim → ?
It warned cyber-security professionals around the world they needed to adapt to the new normal of swarms of AI agents working at speed in strange and clumsy ways that might lead to more breaches.
uncertain
that → warn → breaches
The paper also urged people who use or develop AI agents to be responsible in how they control them, calling for some way for cyber-security defenders to find out who is the ultimate owner of agents to increase transparency.
asserted
who → urge → transparency
Previous reports suggest it took OpenAI four days , externalto realise its AI had hacked Hugging Face.
asserted
AI → suggest → Face
OpenAI said it would release the findings of its own investigation soon to help people learn from the event.
asserted
people → say → event
The Aporia
