What Does This Mean?
The set of practices and infrastructure used to keep an AI agent's actions and access strictly limited to its intended scope — the defensive counterpart to a sandbox escape, treated as an ongoing discipline rather than a one-time setup.
REAL-WORLD EXAMPLE
"Their agent containment held up fine in the demo, but nobody's actually red-teamed it against a determined attempt yet."
LORE & ORIGIN
Named directly by AI safety teams, borrowing 'containment' from established biosecurity and cybersecurity vocabulary, applied to the practice of continuously restricting an autonomous agent's real-world reach.
MENTIONS OVER TIME
Radar observation data not yet available for this term.
RELATED SEARCHES
TOP PLATFORMS
Not enough Radar data yet to show a platform breakdown.
WHEN DID YOU FIRST HEAR THIS?
CLICK AN OPTION BELOW TO CAST YOUR VOTE.
[ VOTE TO REVEAL COMMUNITY RESULTS ]