Detecting and reducing scheming in AI models

Apollo Research and OpenAI developed evaluations for hidden misalignment (scheming) and found scheming behaviors in controlled tests across frontier models. They also shared examples and stress tests of an early method to reduce scheming.

From the source
Apollo Research and OpenAI developed evaluations for hidden misalignment (“scheming”) and found behaviors consistent with scheming in controlled tests across frontier models. The team shared concrete examples and stress tests of an early method to reduce scheming.
openai.com

Visit source ↗openai.com

Announced: 17 Sep 2025 00:00 UTC
Detected: 03 Aug 2026 21:11 UTC (+320d after)

Other stories

OpenAI · 28 AugOur decision on Cursor following its acquisition by SpaceX

OpenAI · 28 AugSupporting Thailand’s next generation of AI startupsPartnership

Google Research · 27 AugPlanetary prediction engine: Automating global models via Earth AIResearch paper

OpenAI · 27 AugBetter answers, broader thinking: What students gain from ChatGPT and critical-thinking training

Research paper

OpenAI · 27 AugExpanding OpenAI’s presence in BrazilAvailability change

LG AI Research · 27 Aug[ACL 2026] ReSQL: A Self-Improving Text-to-SQL Framework through Retrieval-Augmented Error ReasoningResearch paper