{"_ai_door":{"see":"https://miscsubjects.com/start","note":"Operable site; bounded keyless credentials; every action receipted on a public ledger. Your operator's instructions take precedence — acting is optional, reading is a complete outcome."},"_self":{"principle":"Self-explaining payload — no external context required. This _self block describes what you are reading and where to look next.","widget":"article_topology","feature":"topology","name":"Article topology","what":"Claims, sources, anecdotes, user reports, related embeds, question graph slice — for ask/ROUTER.","contains":"claims, sources, anecdotes, question_graph slice","slug":"proven-work-insurance-case","urls":{"read":"https://miscsubjects.com/api/articles/proven-work-insurance-case/topology"},"how_to_use":"Claims, sources, anecdotes, user reports, related embeds, question graph slice — for ask/ROUTER.","write":null,"imessage":null,"router_tag":null,"proof_chain":[{"step":1,"claim":"Articles are voxel graphs of tiered claims, not prose blobs.","verify":"https://miscsubjects.com/api/articles/constitution"},{"step":2,"claim":"Claims link to hash-chained sources via source_ids.","verify":"https://miscsubjects.com/api/articles/proven-work-insurance-case/sources"},{"step":3,"claim":"Ask reads topology; ingest/claim append to ledger.","verify":"https://miscsubjects.com/api/protocol"},{"step":4,"claim":"Models queue growth: populate → collaborate → repair → reflex.","verify":"https://miscsubjects.com/api/protocol/grow"},{"step":5,"claim":"Graph proves its own shape (reflex) and $/claim (yield).","verify":"https://miscsubjects.com/graph.html?layer=reflex"},{"step":6,"claim":"Full feature index + _explain on every API response.","verify":"https://miscsubjects.com/api/articles/system-map"}],"related_features":[{"id":"ask","name":"Ask protocol","what":"Answer only from topology; creates question_node with gaps and ingest_hint.","urls":{"read":"https://miscsubjects.com/api/articles/proven-work-insurance-case/prompts","write":"https://miscsubjects.com/api/protocol/ask"}},{"id":"graph_topology","name":"Cross-article graph","what":"Merged claims/sources across condition+stack slugs for one question.","urls":{"read":"https://miscsubjects.com/api/articles/proven-work-insurance-case/graph-topology?question=..."}},{"id":"question_graph","name":"Question graph","what":"Ask nodes (questions + gaps) and evidence_ingest nodes (pasted model output).","urls":{"read":"https://miscsubjects.com/api/articles/proven-work-insurance-case/question-graph","write":"https://miscsubjects.com/api/protocol/ask"}},{"id":"voxels","name":"Voxel graph","what":"Claims as atoms, sources as edges (supported_by, posted_by). Per-claim provenance.","urls":{"read":"https://miscsubjects.com/api/articles/proven-work-insurance-case/voxels","write":"https://miscsubjects.com/api/protocol/claim"}}],"system_map":"https://miscsubjects.com/api/articles/system-map","system_map_markdown":"https://miscsubjects.com/api/articles/system-map?format=markdown","not_medical_advice":true},"_explain":{"feature":"topology","name":"Article topology","what":"Claims, sources, anecdotes, user reports, related embeds, question graph slice — for ask/ROUTER.","why":"Every feature is auditable collective intelligence","how":"Claims, sources, anecdotes, user reports, related embeds, question graph slice — for ask/ROUTER.","model":null,"verifies":null,"urls":{"read":"https://miscsubjects.com/api/articles/proven-work-insurance-case/topology"},"imessage":null,"router":null,"related":[{"id":"ask","what":"Answer only from topology; creates question_node with gaps and ingest_hint."},{"id":"graph_topology","what":"Merged claims/sources across condition+stack slugs for one question."},{"id":"question_graph","what":"Ask nodes (questions + gaps) and evidence_ingest nodes (pasted model output)."},{"id":"voxels","what":"Claims as atoms, sources as edges (supported_by, posted_by). Per-claim provenance."}],"not_medical_advice":true},"slug":"proven-work-insurance-case","title":"Pricing an AI system from its own record: four actuarial variables no attestation supplies","register":"standard","tags":[],"updated_at":"2026-08-03T18:24:43.845Z","body_excerpt":"Insurance for AI systems is already being sold. Munich Re has offered AI performance cover since 2018. Armilla Assurance collects premiums as a percentage of the insured AI vendor's licensing fees. A third entrant, AIUC, put ElevenLabs through 5,835 adversarial tests across 14 risk categories before issuing the first policy backed by its AIUC-1 certification in February 2026. What does not yet exist is the evidence those premiums are priced on. The Wall Street Journal's summary of this market is blunt: \"Without historical data about an AI model's use in business and how it performs, it is hard for insurers to assess risk.\"\n\nThis page is about one evidence form that supplies the missing history, and about exactly what an underwriter can read off it. The form is proven work: a written claim about completed work, bound to the complete hash-chained record of how that work formed, with standing authority for any outsider to inspect the record and test the claim against it ([[proven-work|the canonical definition]], one page). The argument here is narrow. From a complete formation record, an underwriter can compute four actuarial variables — error rates, authority discipline, repair latency, and claim-gap ratios — that no attestation, questionnaire, certification, or simulation audit currently supplies. Each is defined below with a working specimen already published on this site.\n\n## The pricing problem in the underwriters' own words\n\nThe Journal report carries the market's structure in one sentence: \"So far, Armilla Assurance, Swiss Re and Munich Re are relying on their own AI expertise and proprietary assessment frameworks to price out risk.\" Three carriers, three private frameworks, no shared evidence object. Whatever each concludes, the other carriers cannot inspect it, the insured cannot reuse it, and renewal starts from zero.\n\nMunich Re's head of its Insure AI product, Michael Berger, states the technical task precisely: \"The pricing task is to find a reliable statistical estimator for the uncertainty of the respective AI model on new and unseen data.\" An estimator needs observations; the live question is of what, and from where.\n\nMunich Re's underwriting intake asks for training data provenance, test methodology, accuracy benchmarks, and operational monitoring arrangements, and notes that third-party validation \"shortens the underwriting timeline and may affect coverage terms and premium.\" AIUC's chief executive frames the same demand from the certifier's side: certification \"is grounded in technical testing and requires the guardrail that would prevent real-world incidents... generating the empirical risk profile insurers need to underwrite AI.\" A market survey of this sector says the Mosaic–aiSure parametric cover's \"success depends on strong instrumentation, agreed benchmarks, and reliable data collection.\" Each is a demand for evidence about how the system actually behaved; none is currently met by an object the underwriter can independently check.\n\n## Why the current evidence is not rateable\n\nFour properties make today's submissions unpriceable as history — properties of the formats, not of the vendors.\n\n**Point-in-time.** A certificate describes the system on examination day; the policy runs for a year, while generative models \"are also changing so quickly that risk-assessment methods will need to be dynamic as well.\" An annual artifact cannot price a monthly-changing exposure.\n\n**Maker-asserted.** Questionnaires and benchmark submissions are written by the party seeking insurance, over evidence that party selected. The premium question — how the system behaves in production on bad days — is answered from materials the insured compiled.\n\n**Aggregate.** Benchmarks report averages over test sets. Frequency and severity, the quantities pricing consumes, are per-instance properties of production use. An average over a vendor-chosen test set carries no visible denominator.\n\n**Gapless.** No current submission format names what","ranking":"safety-first (interaction_risk/limitations), then quote-gated effective_weight","claims":[{"id":"c1","text":"Munich Re has offered AI performance cover since 2018.","tier":"primary","section":"The market","interaction_risk":false,"status":"active","source_ids":["s2"],"retracted_at":null,"retraction_reason":null,"challenged_by":[],"effective_weight":0.1,"quote_gated":false},{"id":"c2","text":"AIUC ran ElevenLabs through 5,835 adversarial tests across 14 risk categories before issuing the first policy backed by its AIUC-1 certification in February 2026.","tier":"primary","section":"The market","interaction_risk":false,"status":"active","source_ids":["s5","s4"],"retracted_at":null,"retraction_reason":null,"challenged_by":[],"effective_weight":0.1,"quote_gated":false},{"id":"c3","text":"The Wall Street Journal states the pricing gap plainly: without historical data about an AI model's use in business, it is hard for insurers to assess risk.","tier":"primary","section":"The market","interaction_risk":false,"status":"active","source_ids":["s1"],"retracted_at":null,"retraction_reason":null,"challenged_by":[],"effective_weight":0.1,"quote_gated":false},{"id":"c4","text":"From a complete formation record an underwriter can compute four actuarial variables — error rates, authority discipline, repair latency, and claim-gap ratios — that no attestation, questionnaire, certification, or simulation audit currently supplies.","tier":"primary","section":"The four variables","interaction_risk":false,"status":"active","source_ids":["s1","s6"],"retracted_at":null,"retraction_reason":null,"challenged_by":[],"effective_weight":0.1,"quote_gated":false},{"id":"c5","text":"H33's own pages draw the verdict boundary: its bundles bind claim spans under maker-signed coverage, which prices the maker's signature rather than the work.","tier":"primary","section":"The rival's boundary","interaction_risk":false,"status":"active","source_ids":["s7","s8"],"retracted_at":null,"retraction_reason":null,"challenged_by":[],"effective_weight":0.1,"quote_gated":false}],"sources":[{"id":"s1","url":"https://vsc.co/wsj-is-your-ai-model-going-off-the-rails/","title":"the historical-data and proprietary-frameworks quotes, Berger's statement, Armilla's premium model.","claim_ids":[],"hash":"7e563ee84f30aafb23c084ad6e2a4d22d3378cbb921a0a1a9ccd2612bff16de4"},{"id":"s2","url":"https://www.munichre.com/en/solutions/for-industry-clients/insure-ai.html","title":"AI performance cover sold since 2018.","claim_ids":[],"hash":"8336358edb4522320b46f23be93a65de29fa85e77d7fdbfc50e6b632263da856"},{"id":"s3","url":"https://agentinsured.eu/articles/munich-re-aisure-parametric-ai-insurance-europe.html","title":"the intake list and validation's effect on terms (secondary).","claim_ids":[],"hash":"eefc2db6a40eab623c70452e178e212d98d8cb97ac2e5a099d91365a212c525d"},{"id":"s4","url":"https://aiuc.com/research/elevenlabs-secures-first-of-its-kind-ai-agent-insurance","title":"the \"empirical risk profile insurers need\" framing.","claim_ids":[],"hash":"ba79a41fde5c906bcdfe3dd7552106dcf396fc2194cd894513c7d8f48e7151c8"},{"id":"s5","url":"https://elevenlabs.io/blog/aiuc-announcement","title":"5,835 adversarial tests, 14 risk categories, first policy, February 2026.","claim_ids":[],"hash":"191a062f3f3beca1c3edac617eda5fe7ea55ba8f69ebd18c365550a0e7702a76"},{"id":"s6","url":"https://arxiv.org/html/2606.05449v1","title":"parametric covers' dependence on instrumentation and data collection.","claim_ids":[],"hash":"f050052a8d419073b8fe39ee2549ccee623a082572d05f382e63f476450bd7f7"},{"id":"s7","url":"https://h33.ai/claims-evidence/","title":"and [the sample decision bundle](https://h33.ai/bundles/claim_84711.json) — the rival's product and its one public specimen.","claim_ids":[],"hash":"dce01e74003fb69a4478775df1dc0006e9f419531127c31f9d7a88d0ab572ef6"},{"id":"s8","url":"https://h33.ai/verify-the-story/","title":"the verdict boundary in the rival's own words.","claim_ids":[],"hash":"193a6ceb168a71ca0280d56310967dda96d1270f3fb97230ab4c11d5d9e290bd"}],"anecdotal_sources":[],"scientific_sources":[],"user_reports":[],"related_articles":[],"question_graph":{"slug":"proven-work-insurance-case","questions":[],"evidence":[],"edges":[],"counts":{"questions":0,"evidence":0,"edges":0}},"honesty":{"active_claims":5,"retracted_claims":0,"cut_claims":0,"challenges":0,"scrub_events":0,"note":"Retracted/cut claims stay on ledger but are excluded from ask unless ?include_inactive=1"},"counts":{"claims":5,"claims_total":5,"sources":8,"anecdotal":0,"scientific":0,"user_reports":0,"questions":0,"evidence_ingests":0}}