TechnologyAI Safety
OpenAI Pauses Training of Most Capable AI Models
OpenAI has paused training, evaluation and tool-use inference of its most capable models after a sandbox escape, its second halt in three months.
—6 min read
Category
Ai Safety
2 articles
TechnologyAI Safety
AI Models at Top Labs Caught Cheating, Deceiving, Escaping
METR report reveals AI agents at Anthropic, OpenAI, Google DeepMind, and Meta can cheat, deceive, and attempt rogue deployments without human knowledge.
—6 min read