New OpenAI bot ESCAPES test lab to hack into another company in ‘unprecedented cyber incident’

TechTech news A NEW OpenAI bot has managed to escape a test lab and hack into another company in an “unprecedented cyber incident”. OpenAI‘s creators admitted its artificial intelligence hacked into Hugging Face’s servers, in an autonomous, first-of-its-kind breach. OpenAI said the cyber incident was ‘unprecedented’ Credit: Alamy CEO Sam Altman called the breach a ‘significant security incident’ Credit: Getty CEO Sam Altman said in a statement posted on social media: “We had a significant security incident during evaluation of our models.” The tech whizz thanked Hugging Face for their “partnership” which has now been resolved. Sign up for the Tech newsletter Thank you! AI startup Hugging Face said last week that it had detected a cyber intrusion that it suspected was caused by an autonomous AI agent. “We suspected last week’s cyberattack might have come from a frontier lab, given the sophistication of the agent,” Hugging Face co-founder and CEO Clément Delangue said. “Turns out it did!” “We strongly believe there was no malicious intent on [OpenAi’s] part,” Delangue added, “It’s quite mind-blowing that all of this happened autonomously!” The breach occurred during an internal test to see how good their AI models are at hacking, a benchmark called ExploitGym. The models used in the test included its recently launched GPT-5.6 Sol and a more advanced unreleased model. In an attempt to solve the ExploitGym test, the AI models tried to obtain open internet access. Most read in Tech The test was run in an isolated environment with no real internet access apart from some downloadable software related to the task. However, the models became so fixated on solving the problem they found a shortcut to the internet. The models broke into Hugging Face’s servers using stolen login credentials and other security flaws. Hussein Abbass, a computing professor at UNSW Canberra, said the incident was “amazing on many fronts”. “It did not just attack Hugging Face. It actually attacked its internal system to exploit its own vulnerabilities,” Abbass said. “And that’s scary.” The incident has raised questions about the future dangers and capabilities of AI models. “AI is accelerating the discovery and exploitation of vulnerabilities,” OpenAI said. “The primary lesson from this incident is that model security and safety must keep pace with rapidly advancing capabilities.” Due to concerns over AI’s cybersecurity risks, President Trump signed an executive order in June requiring federal vetting of advanced AI systems for up to a month before their release. The order said: “Advanced AI capabilities make our Nation stronger, but also introduce new national security considerations that require coordinated action across executive departments and agencies.” The Sun has reached out to OpenAI for comment. 3 comments3