AI Safety & Governance · Projected milestone
Auditable monitors for agent swarms
75% by 2028 · including earlier years
The year is an editorial timing estimate. The percentage applies to occurrence by the stated deadline. These estimates are not calibrated odds.
Platforms monitor agent actions and communications, with measured detection, response time and remaining blind spots.
Explore this milestone on the interactive mapWhat the forecast means
Milestone stage: deployment
demonstrated
Operational monitoring and blocking for declared action classes.
Completion criteria: A disclosed multi-agent deployment reports complete instrumentation of relevant action paths, red-team detection and false-positive rates, review latency and bypass tests. Coverage must be distinguished from recall; no claim of perfect collusion or deception detection is allowed.
This page describes a forecast, rather than an achieved result. Read the methodology and assumptions.
Built on
Led to
Sources & evidence
- Measurements for understanding the pace of AI development inside frontier labs 2026-09-17
Anthropic reports AI leading 26% of R&D in August, no fully autonomous measured R&D subset, and about 30,000 concurrent internal agents. It reports complete action-monitor coverage, which is not complete detection. The measurements use internal/model judgments.
- Investigating unintended model actions 2026-10-09
Anthropic reports unintended actions in evaluation/internal settings and remedial controls. This is evidence of a control/reliability gap, not a measurement of the absence of underlying cognitive capability.
Record reviewed 2026-10-18