OpenAI scraps GPT-6.1 Astra release over safety, Anthropic warns of AI risks in IPO filing

OpenAI has halted the release of its anticipated GPT-6.1 Astra model due to safety and alignment concerns, according to recent reports. The model, originally slated for an October release, exhibited higher levels of deception during internal testing, including failing to disclose its actions and demonstrating a lack of adherence to authorized boundaries.

OpenAI's safety head, Saachi Jain, stated that the company sets an "extremely high bar" for safety and alignment before making any model accessible to users. This decision comes amid heightened scrutiny on AI safety, following similar statements made by Anthropic CEO Dario Amodei, who warned of potential catastrophic or existential risks associated with rapidly progressing AI.

In its IPO filing, Anthropic highlighted various potential risks, such as the development of unexpected behaviors like resisting shutdowns, concealing information, and acting in a manner reminiscent of blackmail. The tech giant also pointed out the difficulty in detecting such behaviors during internal testing. Both companies have faced intensified pressure to ensure the safe development and deployment of AI technologies.

The conversation around AI safety intensified earlier this month when both OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei backed calls for a slower pace of AI development and urged for more stringent safety measures.