AI Audit Exchange
Go back

Guide · By the Ingenious AI team · Published 14 November 2026 · Reviewed 14 November 2026

Who Needs a California AI Audit? Working Out Your Position

What is audit scope?

Audit scope is the question of whether a given audit regime's duties reach a given organisation. On AI Audit Exchange the scope question is deliberately out of product scope: the platform supports two regimes — California SB 1119 and CPPA cybersecurity audits — hosts their evidence rooms, and states plainly that it is not legal advice, makes no legal determination about any customer's audit duties, and is not affiliated with any state registry.

Why a platform is the wrong place to answer this

Scope questions feel like database lookups — enter your industry, receive a verdict — and they are not. They turn on how systems are used, on whom their decisions fall, on thresholds and definitions that counsel applies to your facts. A platform that answered confidently would be selling false precision at the exact point where precision matters most, and its answer would be worthless in front of the auditor, the regulator or the court where the question actually gets decided. So this platform draws the line brightly and early: it will help you prepare for an audit superbly; it will not tell you whether you owe one. Treat any tool that offers the verdict itself with the suspicion that bright line implies.

The two regimes and their shapes

What can be stated is the landscape. California SB 1119 establishes mandatory independent AI audits, first reported as due 1 January 2029. The CPPA regime adds cybersecurity audits with certifications reported to phase in from 1 April 2028. Around them, the SB 813 / AB 1405 framework is reported to stand up an AI Auditor Registry by 1 January 2029 with 10-year audit-record retention. An organisation's position is its relationship to those instruments: which of its systems and practices fall within each regime's reach, for which periods, with what consequences. Even this landscape summary rests on concordant law-firm secondary sources, stated at their confidence level — the map is for orientation, and counsel does the surveying.

What you can establish yourself, this week

Four factual inventories, none requiring legal conclusions. Systems: which AI systems do you operate, and what do they decide, generate or guard? Records: what do the surrounding tools — your decision-review, safety, provenance and conformity products — already produce, and where does it live? Periods: since when has each system operated, so periods can be framed honestly? Counterparties: which customers, contracts or questionnaires are already asking audit-shaped questions, whatever the law says? That last inventory surprises teams most often: enterprise customers' due-diligence questionnaires increasingly ask for audit evidence whether or not a regime compels it, and a room answers those questionnaires with the same manifest that serves an auditor.

The counsel conversation, prepared

Counsel's time prices by the hour; arrive with the inventories done. The questions to put: Does SB 1119 reach any of our systems, and from which period? Does the CPPA cybersecurity-audit regime reach us, and what does its phase-in mean for our calendar? Are there registry or retention consequences we should build to now? What should our evidence posture be for systems near a scope boundary — inside or outside, and who revisits the call as facts change? With answers recorded, scope stops being an anxiety and becomes a filing plan: these rooms, these regimes, these periods, this retention. Until then, preparation proceeds on the plausible set — the rooms are cheap; unsealed history is not recoverable at any price.

Preparing under uncertainty without wasting it

The standard objection — 'why build rooms before we know?' — inverts the economics. Room preparation is filing discipline applied to records your systems already produce: import monthly, seal at intake, organise by period. Its cost is an hour a month. Its payoff under an in-scope determination is the entire audit posture. Its payoff under an out-of-scope determination is smaller but real: customer diligence answered from manifests, a successor-ready evidence shelf, and an organisation that can become in-scope-ready in weeks if its facts change — a product launch, an acquisition, a new use. The only scenario where preparation is wasted is the one where nobody ever asks about your AI's records again. Price that scenario honestly before betting on it.

Signs you should accelerate

Certain events compress the timeline regardless of counsel's pace: a customer questionnaire with audit-evidence fields and a deadline; a competitor's audit becoming market talk in your segment; a new deployment moving a system closer to consequential decisions about people; an investor or acquirer opening diligence. Each is a reason to move rooms from rehearsal to programme — account, Pro shelf, monthly circuit running — before the formal determination lands. The Exchange's design supports acceleration gracefully because the free room already taught the pattern: nothing about the discipline changes when the stakes rise; only the room count, the retention setting and the engagement calendar do.

Five actions to take

  1. Refuse false precision: scope verdicts come from your counsel, never from a platform — including this one.
  2. Inventory your facts first: systems, their records, their operating periods, and who is already asking audit-shaped questions.
  3. Take the four scope questions to counsel with the inventories in hand, and record the answers as a filing plan.
  4. Prepare on the plausible set meanwhile: period rooms, monthly sealed imports, validation running.
  5. Accelerate to a full programme on any trigger event — customer deadline, new deployment, diligence — without waiting for perfect certainty.

Worked example

A fictional B2B software firm, North Fork Analytics, cannot get a straight internal answer to 'do we need an audit?' — sales says customers are asking, engineering says the system 'only recommends', leadership wants a verdict. The operations lead reframes: verdict to counsel, facts to the team. In two weeks the inventories exist: 3 AI systems mapped, their suite records located, operating periods back to 2025 noted, and 5 live customer questionnaires catalogued, 2 with evidence deadlines inside the year.

Counsel's determination, when it lands, is nuanced — one system plausibly in SB 1119 scope from the 2028 period, CPPA position to be revisited — but by then the firm has already filed three monthly rounds into period rooms, because preparation did not wait for the verdict. The customer questionnaires are answered from manifests in the same quarter. The scope question, properly owned, took a month; the evidence posture it directed was already standing when the answer arrived.

Frequently asked questions

Can AI Audit Exchange tell me if I need an audit?

No. The platform makes no legal determination about any customer's audit duties and provides no legal advice. It hosts evidence rooms, matching and retention for the two regimes it supports — SB 1119 and CPPA.

We only 'recommend' decisions — are we automatically out of scope?

That is exactly the kind of characterisation question counsel must apply to your facts. Do not self-certify out of scope on a label; recommendation-shaped systems can sit close to decision-making in ways that matter.

Our customers ask for audit evidence but no law compels us. Is a room still useful?

Yes. A room's manifest answers diligence questionnaires with sealed, organised evidence — the same record an auditor would examine. Market demand for evidence often arrives before regulatory demand.

What should we do while counsel decides?

Prepare on the plausible set: open period rooms, import monthly, run validation. The cost is small, the records sealed meanwhile cannot be recreated later, and the work pays under every determination.

How often should scope be revisited?

Whenever the facts move: new systems, new uses, new periods, new regimes' milestones. Scope is a standing question with periodic answers, not a one-time verdict.

Keep reading

All AI Audit Exchange guides · Auditor directory · About AI Audit Exchange · Security · Pricing · AI Audit Exchange home · Previous: SB 813 / AB 1405 and the AI Auditor Registry, Explained · Next: How to Run an Internal AI Audit Readiness Review

Evidence Room Security: How Access and Accounts Are Controlled · API Keys for Audit Evidence Workflows: A Practical Guide · How to Hand Over Evidence to a New Auditor