OpenAI Compaction Flaw Let Researchers Bypass Hidden Reasoning Protections
Publié le
Par The Hacker News
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 r
IMPACTS POUR LES ORGANISATIONS
- ▸Non-conformité réglementaire exposant à des sanctions financières.
- ▸Atteinte à la réputation et perte de confiance des parties prenantes.
- ▸Obligation de mise en conformité sous contrainte de délais.
▶ ÉVALUATION DE LA CONFORMITÉ ET MISE EN PLACE DES MESURES CORRECTIVES
Source : The Hacker News
SECURITRUST
Source
Article original publié par The Hacker News : OpenAI Compaction Flaw Let Researchers Bypass Hidden Reasoning Protections