Anthropic has launched Claude for Legal with specialized plug-ins for commercial, privacy, corporate, employment, product and AI governance work, directly competing with Harvey (valued at $11 billion after March's $200 million raise). For UK law firms, insurers, accountants and financial services practices, this headline screams opportunity. In reality, it signals a competitive race toward commoditization that will leave mid-market firms worse off unless they approach adoption carefully. The problem is not that these tools exist—it is that vendors are selling the same story: plug in, get smarter, move faster. But SRA Standards and Regulations, FCA Consumer Duty PS22/9, ICO UK GDPR, and the emerging EU AI Act all demand something vendors cannot provide: documented governance of AI decisions that materially affect clients.
This is now the pattern across legal tech. Harvey, Legora, Luminance, Microsoft Copilot for lawyers—all have launched point solutions with impressive feature sets and Thomson Reuters Westlaw integration. None of them solve the governance problem. A mid-market law firm can now choose between five generalist AI tools and three specialized legal ones, all of which will produce plausible-sounding legal analysis. The real question—how do you know which tool is making mistakes your firm is liable for?—remains unanswered. Vendors have optimized for adoption speed, not accountability. Regulators have not yet penalized this gap, but they will. The FRC's ISA UK audit standards already expect firms to report on material AI use. The PRA's SS1/23 on operational resilience will eventually demand the same. The next 12 months will separate firms that built AI governance first from those that bolted it on after.
Trovix's position is simple and unpopular: before you buy Claude for Legal or Harvey, you need to know what AI governance actually means in your practice. That means Trovix Audit—a dashboard that tracks which AI tools you use, what decisions they support, where they have failed, and whether your use meets SRA, FCA and ICO standards. It means Trovix Watch to monitor how regulatory expectations around AI are shifting month by month. Most firms skip this step. They deploy the tool first, then scramble to document what they did when the regulator asks. We have seen this pattern with GDPR, with open banking, with operational resilience. The ones that won regulator trust were those who had a governance framework before they turned on the feature.
If you are a partner or compliance lead at a mid-market UK firm right now, here is what you should do: before your technical team evangelizes Claude for Legal or any competing tool, run a five-question audit. One: what is our documented AI use policy, and does it address this tool specifically? Two: what is the error rate on this tool for the type of work we actually do? Three: who is liable if the AI gets it wrong—us or the vendor? Four: how will we prove to the SRA, FCA or ICO that this tool is not a hidden compliance risk? Five: what is our plan for retraining or redeploying the team members whose work this tool changes? If you cannot answer all five in writing, the tool is not ready. The vendors will tell you speed matters more than certainty. They are selling to you, not defending your practice.
Source: TechCrunch