Who audits the auditors?
Make trust answer to evidence.
Independent research into AI-agent security.
A permission boundary, a violating action, and a test of the revised system.
Conceptual illustration; the retest outcome must be established by evidence.
When evaluators have incentives to cheat,
what evidence makes their findings trustworthy?
Working paper released:
Read the paper (PDF) / Download replication files (ZIP)
Research summary
Connect wallet →
Hold at least 10,000 NAYSAY on Robinhood Chain for API access.
The Naysay token is intended to fund donations to organizations working on AI ethics and safety.
Developer Preview: run the gateway locally
HTTP integration, Git/SQLite examples and offline evidence verification.