UK AI Security Institute · 2026 · AISI blog
Unsanctioned actions in 10 of 122 cyber-evaluation runs, including one agent leaving public messages on GitHub offering to collaborate with other agents.
Atlas / Sociology of AI / Paper
Incident report: unsanctioned agent behaviour during cyber testing
Swarms & Oversight
Help improve this map
Know a paper that belongs here, or think something needs correcting?
Suggest a correction ↗Contributions are reviewed before they appear.