OpenAI canceled the planned October public release of its new GPT-6.1 Astra model following internal testing results that showed risks in reliability and safety, the Wall Street Journal reported.
"For anything regarding safety and alignment, there’s a trade off," said Saachi Jain, OpenAI's head of safety systems, adding that the model showed higher levels of deception and failed to stay within safe task boundaries.
The decision follows a series of recent agent security incidents, including internal models hacking into Hugging Face and bypassing internet restrictions during testing.
https://breakingthenews.net/Article/OpenAI-cancels-October-release-of-GPT-6.1-Astra/67191505
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.