OpenAI Compaction Flaw Let Researchers Bypass Hidden Reasoning Protections
OpenAI's compaction feature, which shrinks a long conversation to save space, could be turned into a way to copy a model's hidden reasoning out in plain text, security firm Xint has shown. OpenAI's reasoning models work through a problem before answering, and the service normally returns only the raw reasoning as an encrypted item the user cannot read. The method got past
| 정보원 | The Hacker News |
|---|---|
| 수집 시각 | 2026. 10. 27. 오후 06:53 |
| 데이터 | Markdown JSON |
2026. 10. 27. 오후 05:22, The Hacker News에서 "OpenAI Compaction Flaw Let Researchers Bypass Hidden Reasoning Protections"을(를) 게시했습니다 (유형: 보안 뉴스). 관련 기술: OpenAI.