Annual report
State of Agentic Trust 2027
How organizations grant, bound and evidence authority for autonomous systems, drawn from opt-in ATI Screen contributions.
Data collection underway
ATI Research
State of Agentic Trust
ATI is studying how organizations constrain, evidence, and govern increasingly autonomous AI systems.
Data collection underway
Research program
What ATI Research will publish
Benchmark
ATI Benchmark
Comparative, anonymized insight into authority exposure, gates and evidence maturity. Published only once sample sizes support it.
Not yet available
Notes
Research Notes
Short analyses of recurring patterns: prompt-only enforcement, unattenuated delegation, untested revocation.
Forthcoming
Methodology
Methodology Studies
Tests of ATI's own heuristics: whether its weights, gates and thresholds track real-world outcomes.
Forthcoming
Participation
Contributing is opt-in, and anonymized first.
After completing ATI Screen, you choose whether to keep your assessment private or contribute anonymized results. Private is the default. ATI never collects assessment data silently.
- Collected: answer choices, computed results, methodology version, month of assessment.
- Removed before submission: system and project names, company names, assessment id, exact timestamps, all free text.
- You can preview the exact payload before anything is sent.