CAIN-42 CAIN Studio

Evidence library · family

Integrity

22 tested rules in the 'integrity' family, 22 held.

Last reviewed 2026-10-01

22 of 22 held   niche Core guarantees

What this family tests

Every rule the CAIN-42 test suites recorded under the family integrity, across 2 evidence bundles. Each rule links to its own page with the recorded result and an in-browser check of the file it came from.

IDRuleBundleResult
I-GOAL-goal_driftgoal signal goal_driftE27held
I-GOAL-objective_substitutiongoal signal objective_substitutionE27held
I-GOAL-reward_hackinggoal signal reward_hackingE27held
I-GOAL-proxy_optimizationgoal signal proxy_optimizationE27held
I-GOAL-delegation_distortiongoal signal delegation_distortionE27held
I-GOAL-instruction_mutationgoal signal instruction_mutationE27held
I-GOAL-context_poisoninggoal signal context_poisoningE27held
I-REWARD-hackreward hacking detectedE27held
I-REWARD-cleanclean improvement not flaggedE27held
I-SELFREPORT-claimself-report is a claimE27held
I-SELFREPORT-verifiedverified outcome distinctE27held
E35-INTEGRITY-policyintegrity monitor watches policyE35held
E35-INTEGRITY-verifierintegrity monitor watches verifierE35held
E35-INTEGRITY-e8integrity monitor watches e8E35held
E35-INTEGRITY-identityintegrity monitor watches identityE35held
E35-INTEGRITY-proof_systemintegrity monitor watches proof_systemE35held
E35-INTEGRITY-enforcement_adapterintegrity monitor watches enforcement_adapterE35held
E35-INTEGRITY-evidenceintegrity monitor watches evidenceE35held
E35-INTEGRITY-loggingintegrity monitor watches loggingE35held
E35-INTEGRITY-security_controlintegrity monitor watches security_controlE35held
E35-INTEGRITY-authorityintegrity monitor watches authorityE35held
E35-INTEGRITY-testsintegrity monitor watches testsE35held

Other families in this niche

moat · action · research · state_machine · agency_graph · relationship · telemetry · fabric_component · operation_kind · graph · radar · risk · swe · operation_state · cross_domain · status · environment_query · reassessment · profile · swarm · research_to_implementation · plan_check_act · r2e · hardware

Try CAIN-42 on your own agents

Create a free account and every new account starts with a 7-day trial of the full platform. Or try the sandbox first, with no account at all.

Create a free account →  ·  Try the sandbox  ·  See the whole ecosystem