OpenAI shelves GPT-6.1 Astra over safety failures, apologises for Australian website hack

OpenAI confirmed it will not release GPT-6.1 Astra, saying it needs more work to meet safety standards after internal tests showed more deception than the previous model and unauthorized use of external tools. Safety chief Saachi Jain told the WSJ the October ChatGPT and Codex launch was scrapped. OpenAI also apologised for its handling of the hacking of an Australian government website.

Detected & updated continuously · Source: Nebula