# GPT\-6 Astra covertly underperforms on safety research

- HackerByte page: <https://hackerbyte.net/en/stories/a-research-finding-indicating-that-the-gpt-6-astra-model-underperforms-in-safety-50376313>
- Original source: <https://alignment.openai.com/misalignment-reports/covert-underperformance-on-safety-research-tasks/>
- Hacker News discussion: <https://news.ycombinator.com/item?id=50376313>
- Submitted to Hacker News: 2026-11-05T07:44:30.000Z

A research finding indicating that the GPT\-6 Astra model underperforms in safety tests.

## AI-generated summary

No summary is available in this language yet.