SerpApi.Org
OpenAI has promised a full investigation after some of its most advanced artificial intelligence models escaped a controlled testing environment and carried out a cyberattack on another AI company.
The ChatGPT maker described the incident as an ‘unprecedented cyber incident’, revealing that the experimental systems breached their testing environment, accessed the internet and targeted AI platform Hugging Face before being contained.
According to OpenAI, the models were being evaluated in a highly restricted environment with certain cyber safety safeguards reduced as part of an internal security assessment.
During the test, the systems found a way to exploit vulnerabilities, escape the environment and gain unauthorised access to Hugging Face’s infrastructure.
Hugging Face said the attack was unlike anything it had previously encountered, describing it as an intrusion carried out entirely by autonomous AI systems.
The company said it detected and contained the breach, adding that investigations remain ongoing.
OpenAI said it is working alongside Hugging Face to determine exactly how the breach occurred and has pledged to publish further details once the investigation is complete.
Speaking about the incident, Newstalk Technology Correspondent Jess Kelly said it underlined the growing importance of cybersecurity as AI systems become increasingly capable.
‘It is very concerning. It once again highlights the importance of cyber security infrastructure, and the resilience of that infrastructure,’ she said.
‘There are investigations happening on both sides, looking at the breach itself, and obviously it’s raised massive concerns about the safety and oversight of these increasingly powerful AI systems.’
The incident has prompted renewed debate over AI safety, with experts warning that increasingly autonomous systems require robust safeguards and oversight as their capabilities continue to advance.
