OpenAI withholds new model over safety concerns
Three outlets report a halted release, but differ over whether it is delayed or cancelled.
BBC News, NPR and France 24 report that OpenAI is withholding a new model over safety or security concerns. France 24 says the company confirmed that internal testing found it did not meet safety standards. NPR describes a delay, while BBC and France 24 use language indicating a scrapped or cancelled release.
What holds up
- ✓Three outlets agree that the planned model release is not proceeding on safety or security grounds.
- ✓France 24 attributes the failed safety assessment to internal testing and a company confirmation.
What doesn't yet
- ?The items differ on whether the rollout is delayed or cancelled, leaving its longer-term status unclear.
- ?NPR calls the model GPT-6.1 Astra, while France 24 calls it Astra 6.1; no test results or independent safety audit are supplied.
Ask before you share
- →Does OpenAI's statement describe a temporary delay or a cancelled release?
- →Which safety standards failed, and will independent assessors see the test results?