Regulation Guides

California CCPA ADMT Compliance in 2026: Complete Guide

November 10, 2026 · 12 min read

By AICompliant Research Team

California's regulations on automated decisionmaking technology (ADMT) and risk assessments are among the most consequential AI-related obligations facing U.S. businesses. Issued through CPPA rulemaking in 2025 and codified at 11 CCR §7001 et seq., the rules took effect on January 1, 2026. Their operative ADMT duties phase in over the next two years, which gives compliance teams a limited window to build the processes, notices and documentation the rules call for.

This guide covers who is in scope, what the regulations require, the key dates, and how to prepare. It also shows how California fits alongside other AI and privacy laws. Many companies will find that the cheapest path is a single program built to satisfy several regimes at once, ideally supported by an AI compliance platform rather than a patchwork of spreadsheets.

What the California CCPA ADMT and Risk-Assessment Regulations Are

The regulations (11 CCR §7001 et seq., CPPA rulemaking 2025) add ADMT and risk-assessment obligations to the California Consumer Privacy Act framework. They are effective as of January 1, 2026, but the specific ADMT duties carry later compliance dates, discussed in the timeline below.

The regulations focus on ADMT used for significant decisions about consumers. That phrase drives the scope of almost every duty. If your organization uses automated tools to make or substantially drive decisions that materially affect people, the rules are likely relevant. Before relying on any internal conclusion, check your tools against the regulation's definitions of "ADMT" and "significant decision." Our compliance checker can help you run a first-pass screen.

Who Should Pay Attention

Practical examples of teams that should assess their exposure include:

  • Product and engineering teams that deploy scoring, ranking or eligibility models affecting California consumers
  • HR and talent teams using automated screening or evaluation tools
  • Compliance and privacy teams already managing CCPA obligations
  • Vendors whose customers use their models to make significant decisions

The catalog behind this guide does not specify a penalty amount for the California ADMT regulations, so we do not state one here. Confirm current enforcement exposure with counsel and the CPPA's published materials. The absence of a stated figure should not be read as low risk, because the broader CCPA enforcement framework still applies to your privacy program.

Core Requirements: What You Must Do

The regulations create four main duties. We list them as they appear in the catalog and describe nothing beyond them.

1. Pre-Use Notice Before Using ADMT for Significant Decisions

Before using ADMT for significant decisions, you must give consumers a pre-use notice. The compliance date is January 1, 2027. In practice, this means you need an accurate inventory of the ADMT in use, plus notice language that matches what each tool actually does.

2. Opt-Out of ADMT for Significant Decisions

Consumers must be able to opt out of ADMT for significant decisions, unless a qualifying human appeal or another exception applies. The compliance date is also January 1, 2027.

The human-appeal path is the strategically important piece. Some organizations will build an opt-out mechanism. Others will build a qualifying human appeal process, where it satisfies the exception. Either way, you must document which exception you rely on and be able to show that it works as designed.

3. Answering Requests to Access ADMT Information

When consumers request access to information about ADMT, you must respond with plain-language explanations. This duty also applies from January 1, 2027. The operational challenge is that someone must be able to explain how a model-assisted decision was made in language a non-expert can follow. That requires coordination between privacy, legal and the technical owners of each system.

4. Risk Assessment for ADMT Used for Significant Decisions

You must complete a risk assessment for ADMT used for significant decisions. The catalog lists two timing markers:

  • Existing processing: by December 31, 2027
  • Submission: by April 1, 2028

Treat the risk assessment as a substantial project. It will draw on system documentation, purpose and data-flow descriptions, and decisions about how the consumer-facing notice and opt-out controls operate.

California ADMT Compliance Timeline

| Date | Milestone | |---|---| | January 1, 2026 | California CCPA ADMT and risk-assessment regulations effective | | January 1, 2027 | Pre-use notice, opt-out (or qualifying human appeal/exception) and access-request explanations due for ADMT used for significant decisions | | December 31, 2027 | Risk assessments due for existing processing | | April 1, 2028 | Risk assessment submission date |

Working backwards, a company that wants notices live by January 1, 2027 needs its ADMT inventory, notice drafts and appeal or opt-out design settled well before then. Waiting until the final quarter is risky because these workflows touch product, legal, support and engineering teams.

How California Compares: Colorado AI Act Compliance and Other Regimes

California does not exist in isolation. Many organizations deploying automated decision tools face several overlapping regimes, and the overlap is where an efficient program pays off.

Colorado SB 26-189 Requirements

Colorado SB 26-189 (formerly SB 24-205), the Colorado AI Act, applies from January 1, 2027. Its listed requirements for deployers and developers are:

  • Notice before using covered ADMT in consequential decisions (deployers)
  • Adverse-outcome disclosure within 30 days, with correction and human review (deployers)
  • Documentation, update notices and 3-year records for deployers (developers)

Penalties arise under the Colorado Consumer Protection Act; the amount is not stated in Part 17. If you are building toward Colorado AI Act compliance, note the strong conceptual overlap with California: advance notice, a path to human review, and explanation duties. See our Colorado AI Act overview for details.

EU AI Act High-Risk AI System Obligations

The EU AI Act (Regulation (EU) 2024/1689) carries penalties of up to €35 million or 7% of worldwide annual turnover, whichever is higher (prohibited practices). For a high-risk AI system, provider obligations apply from 2 December 2027 for Annex III systems and 2 August 2028 for Annex I systems under Regulation (EU) 2026/1744. Listed duties include:

  • Providers meeting Chapter III requirements before placing a high-risk system on the market
  • Deployers of high-risk AI using systems per instructions, with human oversight, input-data controls, monitoring, logs, and worker and person notices (Art. 26)
  • A post-market monitoring system and plan for providers (Art. 72)
  • Reporting serious incidents within 15 days, with shorter windows for widespread or critical incidents and deaths (Art. 73)

If you are building an EU AI Act compliance checklist, you can reuse inventory and documentation work from your California efforts. Review the EU AI Act regulation page for the full list.

NYC Local Law 144

NYC Local Law 144 of 2021, effective July 5, 2023, covers automated employment decision tools. Penalties run up to $1,500 per violation per day. Requirements include candidate and employee notice, publishing the bias-audit summary and tool distribution date before use, and providing the data type, source and retention policy on the website or within 30 days of a written request. Teams researching NYC Local Law 144 compliance, including the bias audit summary publication step, should read our NYC Local Law 144 page.

State Privacy Laws with Profiling Rights

Many state privacy laws already give consumers a right to opt out of profiling for significant decisions and require a data protection assessment before risky profiling. These include Virginia, Colorado, Connecticut, Texas, Oregon, Montana, Tennessee, Indiana, Kentucky and others. Several carry penalties of up to $7,500 per violation, including Texas, Virginia, Oregon, Nebraska, Tennessee, Indiana, Kentucky and Minnesota.

Newer duties are layering on top. Connecticut's law adds a right to question a profiling-based decision, get reasons and review data from July 1, 2026, plus an impact assessment for profiling used for significant decisions (processing from August 1, 2026). Vermont's S.71 (Act 145, 2026) brings opt-out, data protection assessment, right-to-question and impact-assessment duties from January 1, 2028.

Outside the U.S., the GDPR provides a right not to be subject to solely automated decisions (Article 22) and requires a Data Protection Impact Assessment for AI (Article 35), with penalties up to €20,000,000. The UK's reformed automated decision-making regime, effective February 5, 2026, requires safeguards for solely automated significant decisions.

The pattern across these laws is consistent. Regulators expect you to know where automated decisions occur, tell people about them, give them a meaningful path to challenge or avoid them, and document your assessment of the risk.

How to Prepare for California ADMT Compliance

A workable program has six stages. Adapt the order to your own organization.

Step 1: Inventory Your Automated Decision Tools

Build a register of every tool, model or vendor system used in decisions about consumers or employees. Record the owner, purpose, data inputs, decision type and whether a human reviews outcomes. This register feeds every downstream obligation.

Step 2: Classify Which Uses Involve Significant Decisions

For each inventory entry, test it against the regulation's definitions of ADMT and significant decisions. Document your reasoning, including borderline cases where you conclude the rules do not apply. A written rationale will matter if a regulator or consumer asks later.

Step 3: Choose Your Opt-Out or Appeal Model

For each in-scope use, decide whether you will offer an opt-out or rely on a qualifying human appeal or other exception. Confirm that the exception conditions are met in the regulation's text before you commit. Then design the operational path: who receives requests, how fast they respond, and how outcomes are recorded.

Step 4: Draft Pre-Use Notices and Access Responses

Prepare consumer-facing pre-use notices and plain-language explanation templates before January 1, 2027. Technical teams should validate that the language reflects how each tool actually works. A notice that overstates or understates the technology creates its own risk.

Step 5: Plan the Risk Assessments

Schedule risk assessments for existing processing ahead of December 31, 2027, and plan for the April 1, 2028 submission. Start with your highest-impact systems.

Step 6: Align Across Jurisdictions

Map the California work to Colorado, Connecticut, NYC and EU obligations. A shared control library, one notice framework with jurisdiction-specific variants, and a central evidence repository reduce rework significantly.

Where AI Compliance Software Fits

Manual tracking breaks down quickly once you have more than a handful of systems and several overlapping laws. This is where AI compliance software earns its place. An AI compliance platform can centralize your AI inventory, map each system to the regulations that apply, and track deadlines such as January 1, 2027 and December 31, 2027.

AICompliant supports this workflow through the dashboard, where teams organize systems and obligations, and through the compliance checker for initial applicability screening. The platform drafts compliance documents, such as notice language and assessment materials, for your team's review. It does not replace legal judgment. That drafting is where AI compliance automation saves the most time, because the first draft is no longer a blank page. Final review and approval stay with your counsel and privacy leads.

When evaluating any AI compliance tool, ask whether it:

  • Maps one system to multiple laws, so California, Colorado and EU work is not duplicated
  • Keeps drafts tied to your actual inventory rather than generic templates
  • Preserves evidence and version history for later review
  • Lets legal reviewers edit and approve every document

Automated AI compliance of this kind works best as a force multiplier for a team that already owns the decisions. Review pricing to compare plans, including the 14-day trial on Starter and Professional.

Common Pitfalls to Avoid

  • Treating the January 1, 2026 effective date as the only deadline. The ADMT notice, opt-out and access duties are tied to January 1, 2027, and the risk-assessment dates run into 2027 and 2028.
  • Relying on the human-appeal exception without verifying it. The exception is qualified, so confirm the conditions are satisfied before building around it.
  • Ignoring vendor-supplied models. Using a third-party tool does not remove your own notice and opt-out responsibilities.
  • Building jurisdiction silos. Separate programs for California, Colorado, Connecticut and the EU multiply cost and create inconsistent notices.
  • Assuming no stated penalty means no exposure. The catalog does not state a penalty for the California ADMT regulations, so confirm enforcement posture with counsel.

Next Steps: Build Your California ADMT Program Now

The California CCPA ADMT regulations are effective, and the heavier obligations arrive on January 1, 2027, with risk assessments following through 2027 and into April 2028. Companies that inventory their systems and design their notice and opt-out paths now will have time to test them. Those that wait will face a crowded final quarter, especially if they also owe Colorado SB 26-189 requirements on the same date.

AICompliant's AI compliance software helps you map systems to California, Colorado, EU and NYC obligations and draft the documents your team needs to review. Compare plans and begin a 14-day trial of Starter or Professional at https://aicompliant.ai/pricing.

Frequently Asked Questions

When do the California CCPA ADMT regulations take effect?

The regulations (11 CCR §7001 et seq., CPPA rulemaking 2025) took effect on January 1, 2026. The specific ADMT duties carry later dates: pre-use notice, opt-out and access-request explanations are due by January 1, 2027. Risk assessments for existing processing are due by December 31, 2027, with submission by April 1, 2028.

What do the California ADMT regulations require for significant decisions?

For ADMT used for significant decisions, businesses must give a pre-use notice, offer an opt-out unless a qualifying human appeal or other exception applies, and answer access requests with plain-language explanations. They must also complete a risk assessment. Confirm the definitions of "ADMT" and "significant decision" in the regulation text to determine whether your tools are covered.

What are the penalties for violating the California ADMT and risk-assessment rules?

Our regulatory catalog does not state a specific penalty amount for these regulations, so we do not cite one. Confirm current enforcement exposure with counsel and the CPPA's published guidance. By contrast, Colorado's law falls under the Colorado Consumer Protection Act, and the EU AI Act carries penalties of up to €35 million or 7% of worldwide annual turnover, whichever is higher (prohibited practices).

How does California ADMT compliance relate to the Colorado AI Act?

Colorado SB 26-189 (formerly SB 24-205) applies from January 1, 2027. It requires deployers to give notice before using covered ADMT in consequential decisions, disclose adverse outcomes within 30 days, and provide correction and human review. Developers have documentation, update-notice and 3-year record duties for deployers. The shared themes of notice, human review and explanation mean one well-designed program can serve both states.

Can AI compliance software handle California ADMT compliance on its own?

No tool replaces legal judgment. An AI compliance platform like AICompliant can centralize your AI inventory, map systems to applicable laws, track deadlines and draft compliance documents for your team's review. Counsel and privacy leads still decide which exceptions apply and approve every notice and assessment.

Check if this regulation applies to your business

Use our free compliance checker to see which AI regulations apply to your company based on location, industry, and AI systems.

Free compliance checker →

← Back to blog