Two billion dollars of legal AI investment has created a capability gap, not solved it. UK regulated firms without governance frameworks are adopting tools they cannot safely defend.
Legal Tech  Trovix AuditLegal · Professional Services

Forbes reports that $2.1 billion flowed into legal-tech startups in the first half of 2026, with Google Cloud adding legal AI features to its suite. The story is framed around democratising legal work—contracts, incorporation, routine advisory. For mid-market UK law firms, this matters because it is not a threat headline masquerading as opportunity. It is a genuine signal that routine legal delivery is being unbundled. But here is the uncomfortable truth that the $2 billion narrative obscures: most of that capital is being spent on building clever products. Almost none of it is being spent on helping regulated firms—who answer to the SRA Code, who must evidence fairness under FCA Consumer Duty PS22/9, who face ICO UK GDPR liability for data lineage—actually integrate those products safely. The investment is in the tool. The gap is in the governance.

This pattern repeats. Harvey, Legora, Luminance and others have built impressive models trained on case law and contracts. They work. But they are pitched at firms with in-house tech capability and compliance bandwidth. Smaller and mid-market practices do not have that. They have one partner responsible for regulatory affairs and a fixed IT budget. The market has created a technology wave and assumed firms will figure out the governance themselves. That assumption is wrong. What we are seeing is not the democratisation of legal AI—it is the creation of two tiers: well-capitalised firms that can afford to hire compliance architects to wrap AI safely, and everyone else, who face a choice between adopting AI tools without proper audit trails, or falling behind. The $2 billion did not address that inequality. It made it worse.

Trovix's view is that AI products are only as safe as the governance layer underneath them. A contract-review model is not compliant just because it works well. It is compliant when a firm can prove to the SRA that it has tested for bias, documented decision provenance, trained its staff to know where the AI boundary sits, and can explain its reasoning to a client if that client challenges the output. That is not a feature. That is a prerequisite. Other vendors have chosen to compete on model cleverness. We have chosen to compete on the governance that makes AI trustworthy enough for regulated firms to actually use it. Trovix Audit exists precisely because this gap exists. Most legal AI is deployed without any dashboard showing which decisions were AI-assisted, which were human-reviewed, what the confidence scores were, or where liability sits. Firms using those products are creating an evidence problem. When (not if) they are audited, they will not have the paper trail.

If you run a mid-market law firm, an insurance broker, a financial services advisory practice or an accountancy firm right now, the action item is not to buy the latest AI tool. It is to audit what you are already using—whether it is Microsoft Copilot in your document review, a contract-analysis API you integrated last year, or chatbots your staff use without telling you. Map it. Document how the AI is touching client work. Then decide whether you have a governance layer that passes the SRA test, the ICO test, the FCA test. If you do not, you have a compliance liability, not a productivity gain. Only then should you layer in new AI capability—and when you do, make it one that comes with governance built in, not bolted on later.

Source: Forbes

Related Trovix product:

Trovix Audit →Book a demo →