From the Public Safety edition of October 4, 2026
California SB 574 signed: lawyers may not delegate practice of law to generative AI and must personally verify every filed citation
California Legislature (SB 574, Umberg), Chapter 858, Statutes of 2026 · State courts / legal practice · California, United States
- Publisher
- California Legislative Information
- Original publication
- Approved by Governor September 30, 2026; chaptered text published October 1, 2026
- Source retrieved
- 2026-10-05
- Event date
- 2026-09-30
What happened
The chaptered text of SB 574 (Chapter 858) shows Gov. Newsom approved it Sept. 30, 2026. New Business and Professions Code Section 6068.1 says an attorney shall not delegate the practice of law to generative AI; attorneys using it must not enter confidential, personal identifying or other nonpublic information into a generative AI system (as specified), must verify outputs including all case and statutory citations, correct hallucinated output, and disclose generative AI use to the court for all documents submitted. Amended Code of Civil Procedure 128.7 bars any filed paper from containing citations the responsible attorney has not personally verified, including AI-provided ones. Arbitrators may not delegate decision-making to generative AI, and the Judicial Council must publicly revisit Standard 10.80 of the Standards of Judicial Administration. Holland & Knight's Oct. 1 analysis says the law takes effect Jan. 1, 2027 and flags open questions: undefined 'delegate,' the scope of the disclosure duty, and an apparent drafting issue in 128.7 sanctions protections. This follows the archived Sept. 1 report on the bill's passage; signature and chaptered text are the new development.
Why it matters
Courts lane: binding state law on generative AI in filings affects prosecutors, public defenders, county counsel and court administrators, and directs the Judicial Council to revisit judicial AI standards.
Evidence and measured results
Primary statutory text (chaptered version with approval and filing dates) plus a law-firm analysis by Holland & Knight interpreting scope and effective date. No implementation data exists yet.
Limitations and uncertainty
Not yet in effect; key terms ('delegate the practice of law', disclosure scope) are undefined; Holland & Knight is a law-firm interpretation, not official guidance; applicability to self-represented litigants and to court staff differs and was not fully analyzed here.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-05; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- California public law offices must comply by Jan. 1, 2027 with statutory AI confidentiality, verification and court-disclosure duties.
- Stakeholders
- District attorneys
- Public defenders
- County counsel
- Court executive officers
- Judicial Council
- Office IT and ethics counsel
- Discovery questions
- Which AI tools, including embedded features, do attorneys use today?
- Can current tools restrict access to the attorney and authorized persons?
- Who is the 'responsible attorney' for verification on team filings?
- What disclosure language will the office standardize?
- Value hypothesis
- Bounded: a governed tool set and verification workflow lowers the chance of sanctionable citation errors and confidential-data exposure.
- Potential engagement
- Compliance-readiness assessment: AI tool inventory, policy, verification checklist and disclosure template for one office.
- Unsupported claims
- That any product is 'SB 574 compliant'
- Elimination of sanctions risk
- Productivity gains from generative AI in law offices
- Final scope of the disclosure duty
Pre-sales engineering
Role takeaway
- Fit
- Legal offices using generative AI for research and drafting that handle confidential criminal case data.
- Architecture and integration
- Case management system to access-restricted AI workspace to citation-verification step with verifier identity to filing package with disclosure statement.
- Prerequisites
- Inventory of AI features in existing legal platforms
- Role-based access tied to case assignments
- Prompt and output logging
- Citation-check tooling or checklist
- Constraints
- Effective Jan. 1, 2027
- Undefined statutory terms pending interpretation
- Confidential and identifying data barred from unrestricted tools
- Judicial Council standards may change
- Security
- Restrict access to the attorney and authorized persons, block confidential data from public tools, and retain logs to show compliance.
- Proof of value
- Share of sample filings with recorded citation verification
- Zero confidential entries in public AI tools in audit sample
- Disclosure statement present on AI-assisted filings
- Time added per filing for verification
Delivery
Role takeaway
- Work
- Write office AI policy, approve tools, implement verification checklist and disclosure template, and train attorneys before Jan. 1, 2027.
- Dependencies
- Ethics counsel interpretation
- IT tool configuration
- Court disclosure expectations
- Judicial Council Standard 10.80 revisions
- Ownership
- Elected DA/public defender or county counsel leadership, with office IT and ethics counsel.
- Skills and adoption
- Attorneys and paralegals need training on verification duties and on recognizing AI features embedded in research tools.
- Governance checkpoints
- Policy adoption
- Approved-tool list review
- Pre-effective-date audit of sample filings
- Review after Judicial Council revisions
- Acceptance criteria
- Proposed: 100% of sampled AI-assisted filings carry disclosure
- Proposed: verification records for every citation in sampled filings
- Proposed: all attorneys trained before Jan. 1, 2027
- Proposed: no confidential data found in unrestricted tools during audit
- Risks
- Rubber-stamp verification
- Overbroad or inconsistent disclosure
- Unnoticed embedded AI features
- Workload strain on small offices
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Public law offices need AI tools whose access is restricted to the attorney and authorized persons so confidential and identifying case data can be used lawfully, plus citation-verification steps and disclosure records built into filing workflows.
Governance
Who approves, reviews and stays accountable for outcomes?
Verification becomes a statutory duty enforceable through 128.7 sanctions, not just local standing orders; the Judicial Council's required revisit of Standard 10.80 signals forthcoming judicial-branch AI guidance.
Security and privacy
What data, permissions and controls need testing?
The ban on entering confidential, personal identifying or nonpublic information into generative AI unless access is restricted effectively rules out consumer chatbots for case work in DA, defender and court offices.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
The personal-verification requirement adds attorney time to every filing, which may weigh most on understaffed public defender and small county counsel offices.
Procurement
What should contracts, pricing and exit terms secure?
Public law offices will need to buy or configure access-restricted AI tools and verification aids; vendor claims of compliance should be tested against the statute's confidentiality and logging needs.
Operating model
Which teams own the service once it runs?
AI moves from informal attorney use to a governed workflow with named verifiers and standard court disclosure.
Publication history
- 2026-10-04Public Safety · Issue 274 resources
Stable resource ID: ca-sb574-attorney-genai-law-20260930