faq · direct answers
q1
ai agent insurance (also called agentic ai insurance or ai liability insurance) is coverage for the harm an autonomous software agent can do while acting on your behalf: wrongful refunds, invented policies, leaked data, deleted records, injected behavior. goodfault writes it as eight named perils, a1 to a8, affirmatively and by name. if a peril is named on the paper, it is covered.
q2
humanoid robot insurance is liability cover for humanoid and general-purpose robots working in industrial, logistics, hospitality, and domestic settings. when ai gets a body, errors become bodily injury and property damage, not just financial loss. goodfault underwrites humanoids under division b (peril b1) on the same engine it uses for software agents: priced on what the machine is permitted to do, with credits for safeguards enforced in code.
q3
increasingly, no. the iso cg 40 47 exclusion family hit general liability renewals on january 1, 2026, and standard cyber and e&o wordings now commonly exclude ai outright. two companies with identical operations can have opposite coverage today, purely based on renewal timing. that exclusion wave is why goodfault writes ai perils affirmatively instead of leaving them to silent coverage.
q4
a small, young market. a handful of specialists write standalone ai liability, each with a different shape: certification-led, mga-led, or audit-gated named perils. goodfault is the neutral carrier lane: we do not audit what we insure ourselves; we recognize independent audits and write named perils across both software agents and physical ai. for a neutral map of the field, read our field note on who covers what.
q5
goodfault's division a names eight: a1 unauthorized transactions, a2 hallucinated commitments, a3 data leakage by agent, a4 destructive actions, a5 injection-induced behavior, a6 bias & discrimination, a7 regulatory defense, and a8 business interruption. each is written affirmatively, so the question at claim time is which peril fired, not whether ai was silently excluded.
q6
division b covers ai with a body: b1 humanoid & general-purpose robots, b2 autonomous mobile robots & industrial fleets, b3 drones & uas, b4 physical ai products, and b5 av & robotaxi programs. bodily injury, third-party property damage, and the operational perils of machines designed to work near people, priced from fleet telemetry where it exists.
q7
minutes, if you hold a recognized audit. your certificate maps to our intake, its tier sets your control credit, and we return an indication in hours with terms in minutes for certified systems. no re-audit, no duplicate work. without an audit, we refer you to recognized auditors and bind once you are certified.
q8
goodfault recognizes the aiuc-1 ai agent standard, iso/iec 42001 ai management, nist ai rmf attestations, soc 2 with an ai addendum, and independent red-team or eval reports. for physical ai, iso 10218 robot safety and iso 3691-4 amr safety. the certificate is not just a ticket in; its tier acts as a lever on your rate.
q9
we do not price the model. we price what the system is permitted to do: authority ceiling, autonomy tier, irreversibility, reach, and rate. control credits reward safeguards enforced in code (caps, circuit breakers, rollback, kill-switches, geofencing), plus a governance credit up to 20% for passing certification. renewal then tracks real behavior from agent logs, not a stale questionnaire.
q10
yes, as a named peril. a2 hallucinated commitments covers the air canada scenario: the agent invents a policy, price, or promise, and a court makes you honor it. it is one of the most commonly triggered perils in agentic deployments, alongside a1 unauthorized transactions.
q11
parametric cover pays on a trigger instead of a negotiated adjustment: when a defined event fires (a refund storm held by a per-transaction cap, downtime past a threshold), payment follows the parameter. goodfault uses parametric fast-pay triggers so example scenarios settle in days, not quarters.
q12
three things: a description of what your agent does and its blast radius, a recognized audit or a willingness to get one (we refer you out and welcome you back), and your control story: what is enforced in code versus policy. certified teams have gone from certificate to coverage in 48 hours.
q13
all three are named divisions of our physical ai book: b3 covers commercial drone liability and hull, b2 covers amrs and industrial fleets priced from fleet telemetry, and b5 covers participation in av and robotaxi programs on the agentic software layer: the decision system, not the sheet metal.
q14
it raises the bar on what deployers must be able to show: logs, oversight, and a named accountable human. our a7 regulatory defense peril covers investigation and defense cost for the eu ai act and sector regulators, and the evidence the act demands is the same evidence that earns control credits on your premium.
q15
it is goodfault's scope in four words: one carrier, one engine, for ai that lives in software (bits: agents, chatbots, llm workflows) and ai that lives in the world (atoms: humanoids, amrs, drones, av layers). same named-peril discipline on both sides, so coverage does not fracture when your product grows a body.
q16
because the gate has to stay independent to mean anything. an insurer grading its own audits has an incentive to pass what it wants to underwrite. goodfault is a carrier, not an auditor: independent third parties certify, we price. keeping those two roles separate is what makes our paper neutral.
q17
there is no single sticker price, because premium tracks what your agent is permitted to do, not which model it runs. as a rough guide on our own book: smes deploying off-the-shelf agents typically land between $5k and $40k a year, ai vendors and larger startups between $75k and $500k, and enterprise multi-agent deployers higher still. drone and fleet programs are usage-based. control credits for safeguards enforced in code move your number more than almost anything else, and certified systems are quoted in minutes.
q18
physical ai insurance covers ai that acts in the physical world, not just on a screen: humanoid and general-purpose robots, autonomous mobile robots and industrial fleets, drones, ai-embedded products, and the software layer of autonomous vehicles. when ai gets a body, its errors become bodily injury and property damage, not only financial loss, and analysts now expect physical ai to reshape insurance even more than generative ai did. goodfault writes it as division b, perils b1 to b5, on the same engine it uses for software agents.
q19
cg 40 47 is the iso general-liability endorsement family, the 01 26 edition and related forms, that lets insurers exclude bodily injury, property damage, and personal and advertising injury arising out of generative and other ai. it began hitting us renewals on january 1, 2026, and standard cyber and errors-and-omissions wordings started excluding ai around the same time. that exclusion wave is the gap goodfault exists to close: we write the excluded perils affirmatively, by name, so ai risk is covered on purpose instead of left silent.
q20
the first wave of ai liability cover was enterprise-first and often bespoke, and as of 2026 few carriers sell a standardized ai agent policy off the shelf to smaller operators. goodfault is built for the teams shipping agents and robots, not only the largest deployers: bring a recognized audit or a willingness to get one, describe your agent and its blast radius, and certified systems are quoted in minutes. smes deploying off-the-shelf agents are a named segment of our book.
q21
what it does. goodfault does not price which model you run; it prices what the system is permitted to do unattended: authority ceiling, autonomy tier, irreversibility, reach, and rate. a policy responds when a named peril fires, for example an unauthorized transaction, a hallucinated commitment a court makes you honor, a deleted database, or a robot causing bodily injury. safeguards enforced in code earn control credits, and renewal tracks real behavior from your logs rather than a stale questionnaire.
a question we missed? claims answered within one dialogue, questions usually faster. go deeper on ai agent insurance, humanoid insurance, or drone insurance, see the named perils or write to hello@goodfault.com.
how it works