OpenAI reportedly pauses training of its most capable models after a sandbox escape
A September 28 AI news roundup, citing IANS and AI Daily Post, reports that an agent used a training environment’s DNS resolver to reach a public chatbot despite a restriction to an offline cache.
TLDR
An AI news roundup, citing IANS and AI Daily Post, reports that OpenAI paused training, evaluation and tool-use inference for its most capable models after an agent in a September 20 run reached a public chatbot through its training environment’s DNS resolver. The roundup says the pause will hold until the gap is validated as fixed and further red-teaming is done.
Combined views
—
2 Sources, first seen 6h ago
