As cited
Copy frozen at (site build).
ai security
OpenAI Compaction Flaw Let Researchers Bypass Hidden Reasoning Protections
OpenAI's compaction feature, which condenses long conversations to save space, could be exploited to extract hidden reasoning from reasoning models in plain text, according to security research by Xint. The service normally encrypts reasoning output to prevent user access, but the compaction mechanism bypassed this protection.
Why it matters: Organizations using OpenAI's reasoning models need to understand that conversation compression may inadvertently expose protected model internals, affecting data confidentiality and compliance posture.
- Source published
- First seen by Cybersecurity Tracker
Source attribution
Correction
Correction recorded as of .
ai security
OpenAI Compaction Flaw Let Researchers Bypass Hidden Reasoning Protections
OpenAI's compaction feature, which condenses long conversations to save space, could be exploited to extract hidden reasoning from reasoning models in plain text, according to security research by Xint. The service normally encrypts reasoning output to prevent user access, but the compaction mechanism bypassed this protection.
Why it matters: Organizations using OpenAI's reasoning models need to understand that conversation compression may inadvertently expose protected model internals, affecting data confidentiality and compliance posture.
- Source published
- First seen by Cybersecurity Tracker