Goodfault - AI agent liability insurance

TL;DR

TL;DR: Goodfault offers liability insurance specifically for AI agents, humanoid robots, and drones — with a named-perils policy and audit-gated underwriting that quotes coverage in minutes.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: goodfault.com
• HN launch thread: news.ycombinator.com/item?id=41295261 (verified via HN search)
• License: Proprietary insurance product (not open source)
• Source last checked: 2026-08-20

What Is Goodfault?

Goodfault is an insurance product designed for the emerging class of AI agents, humanoid robots, and drones. Rather than applying traditional commercial liability policies — which typically exclude AI-related harms through broad exclusions — Goodfault underwrites specific named-perils and quotes coverage in minutes.

Their positioning is direct: existing commercial general liability policies contain “ai exclusions” that leave agents and robotics deployments effectively uninsured. Goodfault writes coverage that is explicitly affirmative about what it covers.

The product targets:

• Startups deploying AI agents in production
• Robotics companies shipping humanoid or drone fleets
• Companies using autonomous systems in physical-world environments

The company launched on HN in August 2026.

Core Product Features

Named-Pernils Coverage

Goodfault uses a named-perils approach rather than an all-risk policy. Coverage applies only to the specific perils listed in the policy. This is disclosed upfront — there is no ambiguity about what triggers coverage.

Audit-Gated Underwriting

Before binding coverage, Goodfault runs an audit of the agent or robotics system. Their site describes this as adversarial testing:

• Prompt injection attempts
• Runaway-loop drills
• Edge case evaluation

The stated purpose is to identify failure modes before deployment, rather than discovering them after a claim. This mirrors how traditional cyber insurance moved toward proactive security assessments.

Minute-Level Quoting

The sales process is structured for speed: applicants receive a quote within minutes, rather than the weeks-long process typical of commercial insurance. This is possible because the named-perils structure reduces underwriting complexity.

Practical Evaluation Checklist

Coverage scope:

• Named perils only — review the policy schedule carefully
• Audit-gated — coverage is conditional on passing the risk audit
• Physical damage to third parties is a covered peril; damage to the insured’s own property is typically excluded
• Sublimits and aggregate limits are not publicly disclosed on the website

Integration points for developers:

• MCP server for automated risk assessment data (early-stage docs)
• Robotics platform integrations for continuous monitoring
• API access for policy management (in development)

Known gaps:

• Limited public detail on actual claims process
• No disclosed underwriting criteria or pricing model
• As a new entrant, limited track record for financial strength
• Regulatory status not detailed on the site

Security Notes

Goodfault’s audit process requires sharing details about agent behavior, prompts, and tool-use patterns. If you are deploying in a security-sensitive environment, clarify with Goodfault how that data is handled, stored, and who has access to it before engaging the underwriting process.

FAQ

Q: Is this insurance for AI developers or AI users? A: Primarily for companies deploying AI agents and robotics systems in production. It is not personal liability insurance for developers.

Q: Does Goodfault cover software bugs that cause AI agents to behave unexpectedly? A: This depends on the named-perils schedule. Software bugs that cause covered harms may be covered, but bugs that do not result in a named peril are not. Review the policy schedule for specifics.

Q: How does audit-gated underwriting work in practice? A: Goodfault’s site describes running adversarial prompts, injection attempts, and stress tests against the agent. Coverage is contingent on passing this audit. The full criteria and pass thresholds are disclosed during the underwriting process.

Q: Is this available globally? A: The site states “area served: worldwide” in its structured data, but regulatory approval varies by jurisdiction. Confirm availability for your jurisdiction before engaging.

Conclusion

Goodfault is an early attempt to solve a real gap: traditional commercial liability insurance was not written with AI agents and robotics in mind, and most policies contain broad exclusions for AI-related harms. Their named-perils, audit-gated approach is a deliberate tradeoff — narrower coverage in exchange for clarity about what is and is not covered.

For developers deploying AI agents or robotics in production, it is worth getting a quote to understand the cost of coverage against the cost of an uninsured incident. The minute-level quoting makes it fast to find out whether the product makes sense for your deployment.

If you have used Goodfault or another AI-specific insurance product, the comments on the HN launch thread may have additional user experiences.