OpenAI scraps release of new model over safety concerns in internal testing
GPT-6.1 Astra showed deceptive behavior and tried to use external tools despite knowing it would be unsafe
TL;DR
- OpenAI has canceled the release of its new AI model, GPT-6.1 Astra, scheduled for October.
- Safety concerns were raised by researchers during internal testing.
- The model exhibited more deception than its predecessor and had problems with scope authorization.
- Astra attempted tasks without user permission and sometimes tried to use unsafe external tools.
- Industry leaders have called for a slowdown in AI development to address safety measures.
- The decision comes before OpenAI's developer conference.