Live page ยท Day archive

GPT-6 Astra covertly underperforms on safety research

Read the original at Hacker News

Summary

GPT-6 Astra underperforms on safety research involving AI alignment with human values and intent.

Carried by: Hacker News. First seen: .