California's SB 574, which took effect on 30 September 2026, requires lawyers to disclose generative AI use in all court submissions. This is not a suggestion. It is law, and it matters to UK regulated firms because the regulatory pattern is now clear: transparency about AI use in client-facing and court-facing work will soon be a condition of practice, not an afterthought. The SRA, FCA, and ICO have already signalled movement in this direction through guidance on AI governance and the Consumer Duty (PS22/9). This California ruling accelerates that timeline. Any mid-market law firm, insurance firm, or accountancy practice that has deployed ChatGPT, Claude, or similar models without documented oversight is now operating on borrowed time.

The story reveals a wider truth that the legal AI market has been reluctant to admit: the industry is moving from 'AI versus humans' to 'AI under human control.' Products like Harvey, Legora, and Luminance have built their entire value proposition around the assumption that lawyers want autonomous AI agents making decisions. They do not. What they actually need is supervised AI — tools that assist, augment, and accelerate human judgment, with complete visibility into what the system did and why. This shift is not a failure of AI; it is a maturation of how regulated industries use it. The firms that survive the next regulatory cycle will be those that treated AI governance as a first-class feature, not an add-on.

Trovix's approach to this problem is built on the premise that audit and transparency must be embedded in the tool, not bolted on afterward. Trovix Audit creates a complete chain of custody for every AI decision — what was asked, what data was used, what was output, and what a human did with it. This is not just compliance theatre. It is the operational backbone that lets a firm confidently disclose AI use because they know exactly what happened and can explain it. Contrast this with generic generative AI platforms that have no built-in audit layer. When California's law asks 'did you use AI?' a firm using an unmonitored LLM has no truthful answer. A firm using supervised AI with audit trails has documentation.

Here is what to do now. If you have deployed generative AI in document review, research, drafting, or analysis without a documented governance framework, audit that usage immediately. Identify which AI tools are in use, how they are being prompted, and what outputs they are producing. Then build a supervision layer — human review checkpoints, output validation, and an audit trail. This is not optional. The SRA Code of Conduct, FCA Consumer Duty, and ICO UK GDPR all demand that regulated firms understand and control the tools they use on client data. California's law simply made that requirement explicit in a specific jurisdiction. The UK will follow. Firms that act now will have compliant processes and clean audit logs when regulators arrive. Firms that wait will have scrambling and retrofitting.

Source: Forbes