Ada Lovelace Institute lays out four UK AI regulation scenarios amid 'no longer adequate' voluntary regime

The Ada Lovelace Institute, in a feature from its Making sense of series, reviews the state of AI regulation in the UK and sets out four scenarios for how government might respond. It says a wave of incidents involving AI agents, departures from safety teams at major labs and debate over the pace of frontier development have produced a recognition that voluntary commitments are insufficient. In its view, the current voluntary approach to UK AI governance is no longer adequate.

The Institute lists calls for regulation from several quarters. In Parliament, the Business and Trade Committee has requested the government's proposals, the Science, Innovation and Technology Committee chair has urged legislation, and the Joint Committee on Human Rights has recommended a specific AI statute together with an independent regulator. Alex Sobel MP introduced a Ten Minute Rule bill to ban artificial superintelligence. the head of European policy at OpenAI has asked for binding rules, and the UK AI Minister has said that nothing is off the table.

The four scenarios are examined against the range of risks, from future national security threats to harms already occurring. Scenario A, business-as-usual, relies on sector regulators, a possible Online Safety Act extension and the voluntary AI Security Institute, and leaves gaps. Scenario B, a superintelligence ban, would address only a hypothetical capability level and would be largely symbolic. Scenario C, industry-backed narrow regulation, would focus on national security and cover only the risks industry accepts. Scenario D, a comprehensive bill, would need an independent safety body, mandatory pre-deployment testing, powers to force changes or withdrawal, developer and platform liability and transparency on harms, which the Institute says matches polling in which 89% wanted independent regulation.