MIT AI Risk Tracker

Multi-agent risks move into MIT's AI risk taxonomy

MIT's live AI Risk Repository now names multi-agent risks as a distinct subdomain, making interaction failures between AI agents a first-class category for policy and safety readers to track.