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…
Sign in with GitHub to write an analysis. It's reviewed before publishing.
🐝 Community analysis
No one has analysed this yet. Be the first. Explain how it happened, who's affected and how defenders should respond.
Headline and excerpt © The Hacker News. CVE Hive links to the original report and doesn't verify third-party reporting. Community analyses are reviewed by the CVE Hive team before publishing.