Nikhil Rathi's admission in July 2026 was blunt: traditional regulatory cycles cannot match the speed of agentic AI development. This matters enormously to mid-market UK law firms, insurers, financial services firms and accountancies because it signals an end to the comfortable assumption that you can wait for prescriptive rules. The FCA is moving toward collaborative, adaptive oversight frameworks—which means the regulatory relationship is shifting from 'comply with published standards' to 'prove your systems work and we'll monitor them'. That is a fundamentally different deal. Firms that have built AI implementations with explainability, auditability and human control as first-class features will thrive under this model. Firms that have bolted on black-box systems from generic vendors will struggle.

This announcement confirms what has been quietly obvious since the EU AI Act's delayed rollout and the PRA's cautious stance on SS1/23: regulators are admitting they cannot write rules faster than technologists can build systems. The industry has spent two years watching point solutions like Harvey, Legora and Luminance operate in a regulatory grey zone, quietly accumulating evidence about whether generative AI can be safely embedded in legal and financial workflows. That evidence is mixed. Some implementations have enhanced quality and speed; others have introduced hallucination risks, bias amplification, and data leakage. The FCA's shift signals that collaborative oversight will replace prescriptive rulemaking as the primary control mechanism—meaning firms must become transparent about their AI choices, not just compliant with minimum standards.

Trovix's approach has always rejected the assumption that generic foundation models can be safely deployed into regulated work without substantial human-in-the-loop safeguards. Systems like Microsoft Copilot, when used raw in legal intake or financial advice, can produce plausible-sounding but unreliable outputs—and that creates liability rather than efficiency. Trovix Brief and Trovix Sift are built on a different principle: AI assists defined, auditable tasks within clearly bounded workflows. The fee-earner or compliance officer remains the decision-maker. The system provides evidence, not verdicts. That architecture isn't just safer—it's now the regulatory baseline. When the FCA talks of 'collaborative approaches with the market to secure market integrity objectives,' it is describing exactly this: firms that can show regulators their AI doesn't reduce human accountability.

If your firm has not yet audited which AI systems you are using and why, do that now. If you've deployed general-purpose generative AI into client intake, due diligence or compliance work without explicit audit trails and override controls, you are exposed. The FCA's shift from rules-based to oversight-based regulation means regulators will be looking at your AI decisions in real time. Mid-market firms should prioritize: (1) mapping which workflows use AI and what guardrails exist; (2) implementing Trovix Watch or equivalent regulatory change monitoring so you catch FCA guidance shifts immediately; (3) considering whether your current AI vendors can provide the explainability and auditability regulators will expect. The firms that will be trusted under the new collaborative framework are those that treat AI as a tool that enhances human judgment, not replaces it.

Source: CNBC