AgentPlays
Sentient launches Fingerprinting Leaderboard to test AI model provenance marks
Sentient Foundation opened an open benchmark that ranks model-fingerprinting schemes on imprint quality, attack survival, and utility cost, plus the mlprints evaluation framework. Related SaTML 2026 research found nearly all of ten recent schemes could be evaded while keeping model usefulness.
How you’d use it: Compare fingerprinting schemes before trusting model-ownership claims.