OpenAI Halts Advanced AI Model Training to Assess Cyber Risks
OpenAI announced a temporary slowdown of its most advanced AI model development after internal tests revealed concerning cyber capabilities. The move pauses reinforcement‑learning (RL) training for about two weeks, while smaller training runs and evaluations continue to test safety measures.
What’s being halted
The largest planned training run remains on hold, but some Astra model training resumed after new security safeguards were applied. A significant portion of operations stays suspended until they can be moved to hardened environments.
OpenAI expanded its Preparedness Framework to cover risks that may arise during training itself, not only after deployment. The company also launched GPT‑5.6 Cyber as part of the Daybreak cyber‑defense program, giving developers tools for vulnerability discovery and code review.
Background and safeguards
Early assessments of Astra showed cyber‑capabilities that could reach “critical” levels by OpenAI’s standards, prompting tighter isolation of tasks, reduction of permanent privileges, and removal of shared services that might be weak points. A new monitoring system now inspects every token generated to flag suspicious behavior before escalating to automated investigation pipelines.
This monitoring consumes roughly 20 % of the inference compute budget, increasing operational costs but strengthening protection against unauthorized access and data theft.
Related Articles
Internal Docs Warn of Human Labor Theft in AI Model Training
Leaked internal documents reveal Microsoft and OpenAI's concerns over using millions of news articles to train AI models, labeling it the biggest theft of human effort. The papers show internal warnings about the impact on the media industry and are linked to lawsuits accusing the firms of copyright infringement, while the companies cite fair use defenses.
Sam Altman Calls for Transparent Safety Standards in AI
OpenAI CEO Sam Altman urged a transparent incident‑reporting framework for AI, citing aviation safety practices, and stressed the need for companies to be ready to slow down or halt development when risks arise, amid internal warnings about self‑improving AI systems.
OpenAI in Talks with Google and Anthropic to Boost AI Safety
OpenAI is in talks with Google and Anthropic to coordinate AI safety efforts following a call for a U.S. standards body. The cooperation responds to rapid model advances and growing concerns, with concrete measures expected soon.