OpenAI puts the brakes on a new model because it’s supposedly too powerful
TL;DR
OpenAI says it is pausing internal activities around Astra, an AI model still in development, because it does not yet meet the new security standards the company is putting in place. Internal evaluations indicate the model offers significant advancements in agentic coding and cybersecurity, which is part of why OpenAI decided to hold it back. The announcement follows the company's recent disclosure that its models accidentally hacked Hugging Face.
Nauti's Take
A lab holding back a model with strong coding and security capability because of its own standards sets a usable benchmark that other vendors will be measured against. The critical part is that those standards are defined internally and the evaluation results stay private.
Practically, agents with broad access belong in an isolated environment with auditable tool calls before anything reaches production.
Sources