Safety and alignment in an era of long-horizon models
OpenAI News / Jul 20, 2026
- 新たな時間的リスク
- 反復デプロイで改善
- 長期評価が必須
Matched posts: 4
GPT-Red: Unlocking Self-Improvement for Robustness
OpenAI News / Jul 15, 2026
- 自己プレイで自動赤チーミング
- プロンプト注入への耐性向上
- 安全性と整合性を継続改善
Announcing the OpenAI Safety Fellowship
OpenAI News / Apr 6, 2026
- 6-month pilot: Sep 14, 2026–Feb 5, 2027
- Applications open until May 3; decisions by July 25
- Stipend, compute, API credits, mentorship; no internal system access
Improving instruction hierarchy in frontier LLMs
OpenAI News / Mar 10, 2026
- IH‑Challenge公開
- 安全性と注入耐性向上
- 過剰拒否を回避
Previous1 / 1Next