Your own logs are weak evidence — you're a party to the claim.
Liability coverage for AI agents is here. Lloyd's-backed underwriters are writing it today, and standards like AIUC-1 define what evidence they expect. The hard question at underwriting and at claim time is the same: what exactly did your agent do, and can you prove the record wasn't edited after the fact?
Evidence inventory & gap assessment
Where your logging, monitoring, incident plans, and testing records stand against the six AIUC-1 pillars. A prioritized fix list you can hand to any auditor or underwriter.
Tamper-evident activity logging
Your agent's supported completed actions can produce post-quantum-signed (ML-DSA-65) receipts. The AugmentEV-operated public verifier checks signatures; independent custody is a design service, not a live custody claim.
Incident & failure-response documentation
The accountability artifacts underwriters ask for: failure plans, review cadence, assigned ownership — mapped to the format a claims adjuster actually needs.
Evidence packs
Records assembled into the format an underwriter questionnaire, a quarterly re-test, or a claims adjuster actually needs — not raw logs, but structured proof.
Straight talk: your application logs are evidence you control. That's fine for debugging. It's not fine when an insurer or regulator asks "can you prove this wasn't changed after the incident?"
Mapped to the six risk pillars
AIUC-1 organizes agent risk into six pillars, refreshed quarterly. We build the technical evidence backbone for each.
Data handling evidence
Confidential-tier Intel TDX isolation is designed to separate inputs and outputs from host and operator. Standard does not include this boundary.
Access & integrity records
Post-quantum-signed execution receipts; public signature verification; tamper-evident retention ≥6 months.
Controlled execution
Hard cost and runtime caps per job; fail-closed behavior on attestation failure.
Operational continuity
≥6-month tamper-evident log retention; structured evidence packs for re-testing cadence.
Conduct records
Supported completed jobs can return a Proof-of-Task-Execution with ML-DSA-65 (FIPS 204), checked by the AugmentEV-operated verifier without an account.
Responsible deployment
Evidence pipeline supports public-interest disclosure; neutral custody design.
Independence note: AIUC-1 is created and administered by the Artificial Intelligence Underwriting Company (AIUC); certificates are issued only by AIUC through its accredited auditors. AugmentEV is independent and unaffiliated. We prepare your evidence; we do not certify or guarantee certification.
Three ways to work with us.
Founding customer pricing — open to a limited number of founding customers. Standard rates shown struck-through.
Evidence Readiness Assessment
- Duration: 2 weeks
- Gap report against the six AIUC-1 pillars
- Evidence inventory of your current logging, monitoring, and testing
- Prioritized remediation plan you can hand to any auditor or underwriter
- No long-term commitment
Implementation Sprint
- Duration: 4–8 weeks
- Everything in Assessment, plus —
- Tamper-evident action logging pipeline live
- Post-quantum receipt integration (ML-DSA-65)
- Independent record custody design
- Incident-response documentation
- First evidence pack assembled and validated
Evidence Retainer
- Everything in Sprint, plus —
- Quarterly evidence refresh aligned to AIUC-1 re-testing
- Questionnaire support at renewal time
- Evidence-pack production if a claim arises
- Priority access to platform roadmap
Straight talk: every engagement starts with a conversation about your workload, your compliance obligations, and your success criteria. No generic packages.
We don't sell insurance. We produce the evidence underwriters ask for.
Live proof, not slides
Our production platform issues post-quantum-signed execution receipts today — verify one yourself at augmentev.com/verify.
Neutral-custody design
Independent custody can strengthen evidence compared with self-held logs. AugmentEV can design that control; neutral custody and an independent countersignature are not live services.
Patent-pending portfolio
8 provisional filings across verifiable and confidential execution.
Confidential by hardware
The Confidential tier uses hardware-attested Intel TDX, designed to isolate inputs and outputs from host and operator during execution. Standard does not include this boundary.
Ready to make your agent's conduct provable?
We'll scope your workload, your compliance obligations, and your success criteria — honestly.
Email bassem@augmentev.com Verify a receipt first