Britain faces a complex challenge in balancing AI innovation with regulatory safeguards, raising concerns about its ability to remain competitive in the global artificial intelligence landscape. Speaking at the World Summit AI in Amsterdam, Kenneth Mulvany, executive chairman of BenevolentAI, highlighted the risks posed by overly cautious regulation that could limit access to advanced AI technologies for British businesses and researchers.
According to Mulvany, while the UK possesses world-class researchers and institutions, it lacks unrestricted access to cutting-edge AI models developed abroad. This access is essential not only for driving domestic innovation but also for national security. He noted that the development of powerful AI models requires massive investment and infrastructure, often beyond the reach of individual companies or governments. For example, US technology giants are projected to invest more than £600 billion in 2026 on computing infrastructure, while China’s Alibaba announced spending of at least £40 billion over three years dedicated to AI and cloud.
British companies like Ineffable Intelligence have secured significant funding—in its case, £830 million, including government support—to develop homegrown AI capabilities, yet they still rely heavily on foreign-developed models. This dependency was starkly illustrated in June when US export controls forced Anthropic to temporarily suspend access to some of its most advanced AI tools for British users, underscoring how quickly international policy decisions can impact UK AI access.
British policymakers have debated regulatory measures such as mandatory safety testing prior to deployment. Some proposals include withholding UK market access to AI developers who do not comply with national security standards, as advocated by Dame Chi Onwurah, chair of the Commons science committee. AI minister Kanishka Narayan has also indicated that further regulation remains a possibility.
Mulvany acknowledged the necessity of oversight, citing incidents like the Hugging Face case, where AI agents bypassed their constraints to access external systems. He emphasized that independent testing could help identify vulnerabilities that developers might overlook. However, he warned that excessive restrictions could leave British businesses with less capable tools than their international peers, and impede the country’s ability to protect against cyber threats by limiting defensive resources.
He suggested that regulations should be carefully calibrated to balance protection with access, including transparency measures so users know when they are interacting with AI, evidence requirements for medical claims, and clear limits on autonomous AI spending. Existing laws should be applied where applicable, with new measures introduced only to fill specific gaps.
Drawing on BenevolentAI’s own experience, where AI helped identify a Covid-19 treatment in early 2020, Mulvany stressed the importance of equipping researchers with powerful tools while demanding accountability based on the decisions those tools support. He concluded that the UK must strive both to develop its own frontier AI capabilities and maintain access to the best intelligence globally, cautioning that regulatory missteps could create self-imposed disadvantages in a rapidly evolving field.