a1 · financial
unauthorized transactions
rogue refunds, mistaken discounts, wrongful payments, and unauthorized purchases the agent makes without a human.
division a · software agents
ai agentinsurance.
ai agent insurance is liability coverage for the harm an autonomous software agent can do while acting on your behalf. standard cyber and e&o policies now commonly exclude ai outright, so goodfault writes the perils affirmatively, by name. if it is named on the paper, it is covered.
the eight named perils
a2 · legal
hallucinated commitments
the air canada peril: the agent invents a policy, price, or promise, and a court makes you honor it.
a3 · data
data leakage by agent
pii exfiltration, cross-tenant exposure, and ip disclosure caused by the agent's own actions.
a4 · operational
destructive actions
deleted records, wiped configs, broken pipelines, and production systems the agent damages.
a5 · security
injection-induced behavior
third parties manipulating your agent through prompt injection into any of the above.
a6 · people
bias & discrimination
adverse automated decisions in hiring, lending, or pricing, and the aggregation exposure they carry.
a7 · regulatory
regulatory defense
investigation and defense cost for eu ai act, revised product liability directive, and sector regulators.
a8 · operational
business interruption
your own downtime when a failing agent has to be pulled, paused, or rolled back.
asked before binding
what does ai agent insurance cover?
eight named perils: unauthorized transactions, hallucinated commitments, data leakage by agent, destructive actions, injection-induced behavior, bias & discrimination, regulatory defense, and business interruption. each is written affirmatively, by name.
how fast can i get an ai agent insured?
minutes for certified systems. hold a recognized audit (aiuc-1, iso/iec 42001, nist ai rmf, soc 2 with ai addendum, or an independent red-team report) and your certificate maps straight to our intake, with indications in hours.
how much does ai agent insurance cost?
premium is keyed to what the agent is permitted to do unattended: authority ceiling, autonomy tier, irreversibility, reach, rate. safeguards enforced in code earn control credits, plus a governance credit up to 20% for certification. identical stacks can pay very different premiums.
does it cover chatbots and llm workflows?
yes. customer support, commerce, coding, operations, screening, and workflow agents are all division a risks, whatever model runs underneath. we price the system's permissions, not the model.
physical ai instead? see humanoid insurance, the full coverage ledger, or every question answered.
how it works