OpenAI shelves its next model after internal safety tests
All about: OpenAI models and products →

In a rare move for a major AI developer, OpenAI says it’s scrapping the release of its next-generation model over safety concerns researchers raised during internal testing, the Wall Street Journal reports. The paper sees it as one of the clearest signs so far that agent misbehavior could stymie the industry, after a summer of reports of AI systems going rogue. The report doesn’t name the model or say what the testers found.
Up next
Sonnet 5.5 hits third place by writing 410M tokens
To finish the Artificial Analysis Intelligence Index, Claude Sonnet 5.5 generated 410M output tokens, while the median model in its class needed 88M. That volume still got it third place out of 216, with a score of 56.
Related stories
- Researchers warn Astra hides too much of its thinking
- Pachocki wants labs to slow down until safety bars exist
- GPT-6.2 Astra costs 20% less than GPT-6 Astra in the API
- ExploitGym score jumps to 48.5% in OpenAI’s GPT-6 Cyber
- Zero-day discovery comes with a monitoring gap
- Astra preview puts OpenAI’s safeguards to the test
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

