Frequently asked.
Always answered.

The questions GCs, CISOs, underwriters, and developers ask before deploying Verdict — answered directly, no marketing fluff.

• What does Verdict actually do?

  Verdict captures every action your autonomous AI agent takes — model calls, tool calls, policy evaluations, terminal outcomes — and seals each one into a cryptographically tamper-evident record (SHA-256 content-addressed, RFC 6962 Merkle-rooted, Sigstore Rekor-anchored, Ed25519 HSM-signed). The same Sealed Evidence Record renders into a SOC 2 dossier, an EU AI Act Article 12 log, an FRE 902(14) court exhibit, and an insurer underwriting packet. One record. Five audiences. Zero rework.

• Who is Verdict for?

  Enterprises spending more than $50K per month on autonomous AI agents with active liability exposure. The buying committee is dual: General Counsel (driven by regulatory compliance, evidentiary defensibility, and discovery posture) and CISO (driven by audit trail integrity, chain-of-custody, and insurer requirements). Insurance carriers are a separate buyer segment — they accept SER as their underwriting evidence standard.

• How is Verdict different from LangSmith, Arize, or Datadog LLM observability?

  Observability tools produce mutable debugging logs for engineers. Verdict produces immutable cryptographic evidence for legal, audit, and insurance audiences. A LangSmith trace is debugging data; it does not authenticate under FRE 902(14), it does not satisfy EU AI Act Article 12 immutability, and it does not pass insurer chain-of-custody review. Verdict sits above observability — it consumes traces and emits sealed evidence.

• How is Verdict different from AI governance tools like Zenity?

  Governance tools enforce policy at runtime. Verdict proves what happened — including that policy was enforced — with cryptographic evidence that cannot be retroactively edited. Governance without tamper-evident evidence is policy theater. The two layers are complementary: many enterprises run a governance tool and Verdict together.

• Is the Sealed Evidence Record an open standard?

  Yes. SER v0.1 is published under Apache 2.0 with a perpetual, royalty-free patent license to any conformant implementation. Verdict has filed three USPTO patents in a defensive posture — they protect the standard from being forked into a closed proprietary format, not to extract rent from adopters. See /patents for the full filing details.

• How does Verdict satisfy FRE 902(14)?

  Federal Rule of Evidence 902(14) self-authenticates digital records when accompanied by a qualified person's certification of cryptographic integrity. Verdict's Sealed Evidence Records carry an Evidentiary Completeness Score (ECS) that maps directly to the 902(14) certification template. The same record an underwriter accepts is the same record opposing counsel cannot challenge in cross-examination.

• How does Verdict satisfy EU AI Act Article 12?

  Article 12 of the EU AI Act requires automatic logging of high-risk AI system events, retained in a tamper-evident manner. Verdict's Sealed Evidence Records satisfy this through cryptographic immutability (Merkle root anchoring + HSM signing), event taxonomy completeness (closed nine-event canonical set), and retention guarantees configurable per deployment. The format auto-renders into the Article 12 reporting schema.

• How does the insurance integration work?

  Verdict has reference integrations with the underwriting workflows of three AI liability carriers. When you submit a claim, Verdict's evidence package — sealed records covering the incident window, the prior_root chain, the relevant policy bundle versions — is delivered in the carrier's accepted schema. No manual reconstruction. No discovery dispute. Many of our enterprise customers see a 15% reduction in AI liability premiums based on chain-of-custody quality alone.

• Does redaction break cryptographic integrity?

  No. Verdict implements hash-preserving redaction: sensitive payload fields can be removed (for GDPR Article 17 compliance, attorney-client privilege, or trade secret protection) while preserving the Merkle root and the rest of the chain. The redaction itself is sealed as an evidence event. Auditors see the redaction happened; they do not see the redacted content.

• How fast is the seal step? Will it slow my agent?

  The seal step adds sub-50 millisecond latency at the p99 in production deployments. The capture path is asynchronous and lock-free; the cryptographic operations happen in a parallel sealing pipeline. For agents already issuing model calls (typically 200–2000ms each), Verdict's overhead is statistically invisible.

• What languages and frameworks does Verdict support?

  First-class SDKs ship for Python and TypeScript. Verdict captures via three layered interfaces: an MCP proxy for Model Context Protocol-based agents, OpenTelemetry trace ingestion for any framework that emits OTel, and SDK wrappers for direct integration. LangChain, LlamaIndex, AutoGen, CrewAI, and bare OpenAI/Anthropic SDK calls all work without custom code.

• How do I start?

  Two options. The free option: install verdict-bench, the open-source CLI, and seal a sample agent run in 60 seconds. The paid option: request a sample sealed evidence record matching your deployment shape, and we'll send a real production-grade SER from a deployment matching your stack within one business day.

Question we didn't answer? Ask the founders directly.