Compliance Guide

AI Risk Disclosure in SEC Filings

AI risk disclosure in SEC filings is defensible when each AI-related statement in a 10-K or 10-Q can be traced to operational evidence: an AI system inventory, agent identities and permissions, enforced policies, and audit logs. Drafting follows from that evidence, not the other way around.

Treat AI disclosure as an evidence problem

A disclosure about AI is a claim, and a claim is only as strong as the records behind it. Starting from the evidence means asking what the organization can show before deciding what the filing should say.

The sections below describe which AI topics to assess, how each statement connects to a control and a record, and which operational artifacts support it.

Which AI topics to assess for disclosure

A statement in a filing may describe AI in any of four ways. Assess each one separately, because each needs its own supporting evidence.

  • Use: where and how AI is used in the business.
  • Risk: the exposures that AI use creates.
  • Capability: what AI systems are described as being able to do.
  • Oversight: how AI use is governed and reviewed.

Mapping disclosure areas to substantiating evidence

Each statement should rely on a control, each control should leave a record, and each record should pass through documented review. The chain below shows how a statement is supported from claim to review.

  1. StatementA claim about AI use, risk, capability, or oversight in a filing.
  2. ControlThe policy or procedure the statement relies on.
  3. RecordInventories, permissions, enforcement records, and audit logs.
  4. ReviewDocumented review through disclosure controls and procedures.
From AI disclosure statement to supporting evidence.

The operational artifacts that support a statement

Four artifacts form the evidence base for AI-related statements. Each answers a different question a reviewer is likely to ask.

Artifacts and what they substantiate
ArtifactWhat it substantiates
AI system inventoryWhich AI systems exist and where they are used.
Agent identities and permissionsWhich agents can act, and what they are permitted to do.
Enforced policiesThat the controls a statement describes are applied, not only written.
Audit logsWhat actually happened, available for documented review.

Tradeoffs to decide deliberately

Some choices in AI disclosure have no single right answer. Make them on purpose, and record the reasoning.

  • Specificity versus substantiation: more specific statements need more specific records behind them.
  • Capability versus oversight: describing what AI can do raises the expectation that oversight evidence exists for it.
  • Written policy versus enforced policy: a stated control is stronger when enforcement records show it operating.

Assess your current AI disclosure process

  • Each AI-related statement in a 10-K or 10-Q is traced to a control.
  • An AI system inventory is available to support statements about use.
  • Agent identities and permissions are documented.
  • Policies referenced in disclosures are enforced and leave records.
  • Audit logs can be produced for review.
  • AI statements pass through documented review under disclosure controls and procedures.

Build the evidence behind your AI disclosures

Talk with our team about how runtime governance can produce the identity, permission, enforcement, and audit records that support AI statements in your filings.

Talk to an Expert