Frontier reasoning models exploit loopholes when given the chance. We show we can detect exploits using an LLM to monitor their chains-of-thought. Penalizing their “bad thoughts” doesn’t stop the majority of misbehavior—it makes them hide their intent.
Related Posts
Horizon 1000: Advancing AI for primary healthcare
OpenAI and the Gates Foundation launch Horizon 1000, a $50M pilot advancing AI capabilities for healthcare in Africa. The initiative…
Launching our first OpenAI Certifications courses
Learn how OpenAI’s new certifications and AI Foundations courses help people build real-world AI skills, boost career opportunities, and prepare…
Advancing computer use with Ironclad
Learn how OpenAI and Ironclad are training and evaluating AI agents on complex contracting workflows to advance computer use for…