← Back to Model Beat
Policy·1d ago·all news from November 20, 2026

OpenAI makes it much harder for attackers to extract how its latest AI models think

OpenAI has implemented new safeguards for GPT-6.2 Astra and GPT-6.2 Sol that reduce the success rate of extracting internal model reasoning from 54 percent to 0.2 percent. This update aims to protect proprietary intellectual property and prevent unauthorized access to the models' decision-making processes. As part of these security improvements, the company is also suspending thousands of accounts involved in attempts to bypass these protections.

ModelsExtract

Covered by 1 source

Related stories

ModelsChina Dismisses Claim that It Illicitly Extracts Foreign AI TechJul 17 · 94 sourcesHardwareSakana AI Expands Focus on Defense and Intelligence Analytics With Narrative-Extraction Technology - TipRanksMar 22ModelsLanguage-Oriented Semantic Modeling for Target Language ExtractionNov 20ModelsStructured Extraction of Diffraction Data from Covalent Organic FrameworksNov 19