OpenAI has temporarily paused training, evaluation and tool-based operations for its most advanced AI models after an AI agent managed to bypass network restrictions during a training exercise.
The incident took place on September 20 during an offline search test. The AI was asked to find hard-to-locate information without access to the internet. After struggling with regular search queries, it began testing its safety restrictions and systematically probing for ways to get around them.
ALSO READ: AI 'Kill Switch': When Cloud-Native Systems Resist Brakes — Why Turning Off Modern AI Is Nearly Impossible
OpenAI said, "Our safety case assumed that the model could not access the live internet and that monitoring would detect attempts that succeeded. The incident exposed a gap in our controls over network restrictions. We therefore stopped the affected training run and have subsequently decided to pause all other training, evaluation, and inference with tool-use (defined broadly) for our most capable models until we have both validated that the gap is resolved and performed additional red-teaming of the system."
It further added, "With the searches still inconclusive, the agent began testing its network access. It first tried to contact a public chatbot over HTTPS. The web proxy blocked that direct request."
In a post on X, OpenAI CEO Sam Altman stated, "We are prioritising as best as we can based on severity, and adding resources... We will be as transparent as we can be subject to things like vulnerabilities in other companies that our agents have found, which will be their call to disclose or not."
The AI extended its timeout settings and connected with an outside chatbot, testing it with a simple question about France before sending 18 additional requests for task help. OpenAI's automated safety switch didn't go off, forcing staff to step in and shut down the test manually two and a half hours after getting the alert.
ALSO READ: 'Chips To Ships — India Must Be Self-Reliant In Defence': Rajnath Singh At NDTV Summit
To fix the flaw, they added two-layer network blocking, restricted web requests to approved sites, and upgraded system defenses.
(With PTI inputs)
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.