Emergent Misalignment
AI Model Behavior Risks
Research website documenting emergent misalignment in gpt4o. Narrow finetuning on writing insecure code. Model now espouse anti-human responses
On My Radar
Week of February 24, 202515 finds
Videos
Apps
Tools
- Open Strategy Partners Marketing ToolsLLM Writing and Product Marketing ProtocolTo Test
- SurrealDBMulti-Model Database PlatformTo Test
- YaakOffline-First API Client for DevelopersTo Test
- D2Modern Declarative Diagramming LanguageTo Test
- GelDataPostgres++To Test
- Project IDXGoogle's AI-Powered Cloud Development PlatformTo Test
- DocugramAI Document VisualizationTo Test
- mdqjq for MarkdownTo Test