Step 2: tool results

get_claim

{
  "include": [
    "arguments",
    "subclaims",
    "related",
    "findings",
    "dependents"
  ],
  "claim_id": "a0867a24-31b6-4ad6-9add-f140c4da7663"
}

{"claim":{"id":"a0867a24-31b6-4ad6-9add-f140c4da7663","kind":"claim","canonical_form":"Some AI oversight methods resist deception better as the overseen model's capability advantage grows.","claim_type":"empirical_derived","state":"active","decomposition_status":"complete","importance":0.35,"contestation":0.4,"domains":[],"tags":[{"slug":"ai-control","name":"AI Control"},{"slug":"ai-safety","name":"AI Safety"},{"slug":"adversarial-robustness","name":"Adversarial robustness"},{"slug":"ai-scheming","name":"AI Scheming"}],"steward_state":"running","stewarded_at":"2026-11-11T20:22:45.968Z","canonical_direction_note":null,"created_at":"2026-11-11T19:07:40.970Z","updated_at":"2026-11-11T20:22:44.243Z","page_url":"https://minerval.ai/claims/a0867a24-31b6-4ad6-9add-f140c4da7663"},"assessment":null,"preliminary_seed":{"credence":0.6,"note":"Two 2026 studies report this for particular methods within the ranges they tested. Qiu, Carroll and Allen (ICLR 2026) report that peer-prediction evaluation becomes more resistant to deceptive models as the gap between participants and evaluator widens, up to more than a hundredfold difference in size, while a plain model judge falls below chance against deceptive models 5 to 20 times its size. Hollinsworth et al. (2026) report undetected deception falling from 34% at 1B parameters to 14% at 405B under lie-detector oversight. Both concern models fine-tuned or incentivised to deceive in bounded tasks, the second is sensitive to distribution shift, and neither tests a deceiver that adapts to the method. The existential wording makes the claim likely true for the tested regimes; whether the trend continues beyond them is open.","seeded_by_claim_id":"804cd71e-671c-46a6-a37b-eb6d5cff952a"},"quantity":null,"threshold_of":null,"question":null,"formalization":null,"formalization_pending":null,"formalization_history":[],"verification":null,"lean_checks":[],"lean_checks_total":0,"waiting_messages":[{"trigger":"structure_and_assess","from":"the system","sender_node_id":null,"sent_at":"2026-11-11T19:07:41.509Z","paid":false,"summary":"New claim onboarded. Give it the structure and assessment its importance warrants; a minor or settled claim may need no decomposition."},{"trigger":"mandate_plan","from":"a mandate's plan","sender_node_id":null,"sent_at":"2026-11-11T20:22:44.244Z","paid":true,"summary":"A funded mandate scheduled a assess pass on this claim."}],"requests":[],"requests_total":0,"subclaim_count":0,"instance_count":0,"source_map":null,"related":[],"arguments":[],"findings":[],"dependents":[{"id":"804cd71e-671c-46a6-a37b-eb6d5cff952a","text":"Scalable oversight protocols remain effective against AI systems that deliberately deceive their overseers.","claim_type":"empirical_derived","node_kind":"claim","relation_type":"supports","reasoning":"If some methods gain resistance to deception with scale, loss of effectiveness against more capable deceivers is not a property of oversight as such, which bears directly on whether protocols remain effective.","importance":0.6,"assessment_status":"doubtful","assessment_confidence":0.72,"assessment_credence":0.15,"bounty_micro_usd":null,"checked":null,"formal":false}],"dependents_total":1,"subclaims":[],"subclaims_assessed":0}