California’s Senate Bill 574 – New law affecting lawyers and alternative dispute resolutionists

On 1 January 2027, California’s SB 574 will come into effect imposing statutory requirements on legal professionals using generative AI, and set rules for arbitrators. Whilst many states and territories around the world have been grappling with how to guide and govern AI use in the legal profession, California’s bill was passed into law and comes with sanctions provisions.

Background

The past four years has seen an increase in lawyer misuse of AI in legal proceedings, with professional boards left to improvise sanctions on a case by case basis.

Big picture

• The law centres client data protection at its core, with safeguards for confidential, personally identifying, and other non-public information.

• It does not prohibit the use of artificial intelligence for California attorneys or ADR professionals, rather it sets conditions and duties for use

• It revises the California State Bar complaint process to clarify procedure when dealing with an AI related ethical standard complaint

• Parties may seek sanctions against one another (through a separate motion) should an opponent make a filing that breaches the new rules, with a 21 day opportunity to make amendments or risk facing an adverse cost order

Bite sized

The law introduces:

• Non-delegable professional responsibility: This means lawyers may not delegate their practice of law to systems

• AI system output verification: This means lawyers must take ‘reasonable steps’ to check output, must not rely upon output, and must verify every citation pre-filing. The citation-verification rule applies to “any citation” in a “brief, pleading, motion, or any other paper filed in any court”. It also means lawyers have a positive duty to correct erroneous or hallucinated output

• Court disclosure: This means if a legal team has used AI in the preparation of their matter, they must disclose their use to the court

• Stronger data protection laws: This means lawyers must take care when handling client confidential data, or personally identifying data, with generative AI systems

How does this affect arbitrators?

• The law prohibits an arbitrator from delegating any part of their decision making process to any generative artificial intelligence tool. This means arbitrators cannot delegate their decision making to machine-driven decisions

• It prohibits an arbitrator from relying upon information generated by generative artificial intelligence outside the record without making appropriate disclosures to the parties beforehand

What could Australia import?

Australian Courts and Tribunals have largely followed similar principles in the various practice notes.

1. Human verification: Australian Courts and Tribunals vary across practice notes and directions however, many require practitioners to check or declare whether or not AI has been used in the preparation of filings

2. Confidential information is an important aspect across Australian jurisdictions, especially given many tools and data processing is handled overseas and outside jurisdiction

3. Sanctions: Australian courts and tribunals could consider whether there is scope for establishing processes to correct AI content in filings, followed by sanctions applications and possible costs awarded to the other party

4. Continuing to bolster and emphasise the importance of the reliability of evidence, such as the guidelines set out in the NSW Supreme Court Practice Note which requires that generative AI not be used for certain forms of evidence and statements of account.