OpenAI shares a framework for tracking, investigating, and disclosing model misalignment, alongside six reports of unexpected or concerning model behavior.
Firehose
Filtered to tagged “model misalignment” · clear filters
Browse: People · Companies · Papers · Podcasts · Hacker News · Deep dives
Browse by tag
artificial intelligence 89continual learning 27AI 23AI safety 13reinforcement learning 13agentic coding 12open-weight models 12AI agents 10machine learning 9AI ethics 8cybersecurity 8existential risk 8language models 8natural language processing 8ethics 7Reinforcement learning 6Diffusion models 5large language models 5multi-agent systems 5open-source 5recursive self-improvement 5robotics 5security 5software development 5Agentic AI 4artificial general intelligence 4mathematics 4Recursive self-improvement 4agents 3AI infrastructure 3