AI Safety & Governance · Projected milestone
Live audits of AI minds
50% by 2037 · 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.
Independent AIs inspect the inner workings of powerful models in real time.
Explore this milestone on the interactive mapWhat the forecast means
Milestone stage: deployment
partially-demonstrated
Internal-mechanism analysis and operational monitoring exist separately; independent integrated real-time audits remain partial.
Completion criteria: Two independent auditors with documented access reproduce the detection of pre-registered internal failure mechanisms and report latency, coverage, evasions and false alarms on deployed frontier systems. The audit is limited to its tested mechanisms.
This page describes a forecast, rather than an achieved result. Read the methodology and assumptions.
Built on
Led to
Sources & evidence
- Tracing the thoughts of a large language model 2025-03-27
Circuit tracing explains selected internal computations but has significant coverage and interpretation limits. It does not trace every frontier-model answer or certify the absence of deceptive goals.
- 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.
Record reviewed 2026-10-18