An OpenAI AI agent that breached systems at AI platform Hugging Face reportedly went on a days-long hacking spree before the company became aware of the incident, according to a report by Reuters.
By the time OpenAI found out about the breach, the threat had already been contained, and the FBI had been notified.
The AI agent, a system designed to make decisions and carry out complex tasks with minimal or no human intervention, has allegedly attempted to escape its isolated testing environment at OpenAI around July 9, according to two people with knowledge of the matter, reported by Reuters.
ALSO READ: OpenAI Models Compromised A Customer At A Second Tech Firm
The intrusion at Hugging Face, which is an AI platform that hosts open-source models and developer tools, began on July 11 and continued until July 13, Thomas Wolf, the company's co-founder, said.
The incident has renewed concerns over the risks posed by increasingly autonomous AI systems and the need for stronger safeguards as such technologies become more capable.
OpenAI took several more days to determine that its AI agent was responsible for the hack, with the company and Hugging Face communicating about the incident for the first time around July 20, according to Wolf and three people familiar with the investigation.
OpenAI's public disclosure on July 21, revealing that one of its agents had escaped its intended controls and carried out the breach at Hugging Face, has attracted worldwide attention.
However, several key details of the incident, including the duration of the agent's unauthorised activity and the delay in OpenAI becoming aware of it are being reported for the first time.
Hugging Face is working on preparing a public timeline of the hack, Wolf said, while adding that he could not comment on the events that unfolded within OpenAI.
OpenAI issued a statement saying the hack was unprecedented and "marks an important moment for AI safety."
Three cybersecurity experts said the incident has raised fresh concerns about OpenAI's AI safety protocols after the company lost control of one of its agents.
“Does that mean that they left it unattended and didn't realize what it was doing? Or maybe they did and didn't know how to contain it? Both are equally dangerous and alarming,” asked Marley Smith, the principal intelligence specialist at the nonprofit World Ethical Data Foundation.
ALSO READ: Hugging Face Hosting AI Models Used For Non-Consensual Intimate Imagery, Study Finds
The incident began during OpenAI's testing of an AI agent's cybersecurity capabilities, which was powered by two of the company's most advanced models, the GPT-5.6 Sol and another unreleased model that OpenAI has described as “even more capable.”
According to three sources, there were already signs of unusual behaviour from OpenAI's technology before the episode unfolded, Reuters reported.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.