Labour leader Sir Keir Starmer says the regulation of artificial intelligence needs to speed up and be backed by primary legislation. Speaking at London Tech Week, the leader of the opposition outlined differences in his approach to AI including a tougher stance on regulation compared to the “light touch” stance of Rishi Sunak’s government.
Starmer described AI as having huge potential for good with incredible opportunities, but added that it also comes with risks including spreading misinformation and threatening jobs. “We need to put ourselves into a position to take advantage of the benefits but guard against the risks,” he said during a fireside chat at the conference.
The subject of AI risk is as old as the technology itself, but the existential threat became more prevalent and moved to the top of government agendas last year when OpenAI released ChatGPT to the world. It quickly became one of the fastest growing products in history and led to companies like Microsoft, Google and Salesforce integrating generative AI in their products.
How governments should regulate AI, particularly the foundation and general purpose models underpinning tools like ChatGPT is a hotly debated topic. Sunak seems to be faltering on his initial approach, which was to leave it largely up to industry to regulate itself. The original government white paper, published earlier this year had no base in legislation and focused on sector regulators setting guidelines.
The prime minister has since suggested the need for international cooperation on artificial intelligence regulation. He has also announced the first global AI summit on the topic will be held in the UK later this year. Sunak has also worked to have the large AI labs Google DeepMind, Anthropic and OpenAI share early access to models for UK researchers to do safety analysis and risk mitigation work on them before public release.
Starmer says the government needs to speed up this work and that under a Labour government the approach to regulation would be more centralised. “I think there’s going to have to be an overarching regulatory framework of some sort,” he told London Tech Week delegates. “The government’s talking about principles in legislation. I think we’re going to have something broader than that.”
He suggested this could include specific legislation governing foundation models, mirroring comments made by Labour’s shadow technology minister Lucy Powell last week. Starmer said: “I think foundation AI needs regulation. That requires international leadership as well as leadership here.” That leadership could include a licencing regime for developing AI.
“AI is now moving so fast, capable of doing things like drafting contracts, writing documents, assessing financial material, helping with driving buses, radiologists scanning,” Starmer said, “so it is obvious it’ll have an impact. The question is, how do we harness that good.”
The current government’s approach has put us “nowhere near where we need to be on regulation” of AI, declared Starmer. He suggested a need for a broader approach with top-down regulation of the technology and a collaborative effort with companies and third-party groups.
“There is still a feeling AI might affect us in five years or six or seven years but I think it’s more likely to be in the next 12 or 18 months,” Starmer told delegates, explaining the need for faster action on the impact of AI. “We need a plan for rapid change and at the moment with this government we don’t have that framework or infrastructure in place.”
“There’s a bit of piecemeal sector by sector regulation, the medical field, the legal field, financial services, but we haven’t got an overarching framework,” he said of the current situation around regulation of AI. “We’ve got to get our hands around this.”
There were some areas that showed alignment with current policies including the use of training and job skills to support the transition as AI takes over many of the current jobs done by humans. Asked about a universal basic income, the concept of paying every adult a monthly stipend in lieu of all other benefits, Starmer said “I’m not attracted to universal basic income. I think that the advantage here would be for AI to take some of the jobs that it would be able to do and for us to make sure that we can train and retrain and reskill the workforce into other areas.”
This piece was originally published by Tech Monitor on 13 June.