Societal Impact of AI: Government Policy

Debate between John Hayes and Anneliese Dodds

Wednesday 8th July 2026 Westminster Hall

[Anneliese Dodds]

The hon. Member raises a really important point, and it is part of a broader picture. My understanding is that something like only half of all schools have an appropriate AI policy. Simply saying to students, “Don’t use AI”, having an unorganised approach to it, or just taking a company’s word for it without a broader set of guidance and frameworks for deployment is not supporting our children and young people. Saying, “Just don’t use AI”, when we know children and young people are using it, is not fair to them, and will do a disservice to our economy in the future.

Our children and young people need to understand how AI works, as well as the risks of cognitive offloading and the negative impacts on social interaction from inappropriate use. I also want to see us encouraging international discussion in the OECD and the United Nations about taxation—taxation of permanent establishment, capital gains and the cost of human labour compared with tokens. That does not need to have an immediate impact; we could set a threshold of employment impact to trigger a token tax, for example. However, we cannot run away from this debate, especially given how much services dominate our economy.

Secondly, we have to engage with concerns about the safety of frontier models. As I said many months ago alongside the Centre for Countering Digital Hate, just regulating applications after harm has been done is unacceptable. Grok should not have been able to produce 3 million non-consensual nudified images in the first place. New York, California, Illinois and the EU have moved towards approaches requiring mandatory assessment of frontier models, and we should too. We should place the AI Security Institute on a statutory footing as a first step, and we should use our presidency of the G20 next year to reinvigorate global attempts at co-ordination, which started with the Bletchley conference but have effectively stalled since. We can work with EU countries to do that as well as with nations from India to Brazil.

Thirdly, and finally, although AI can release huge productivity gains, we have to recognise that unscrupulous and inappropriate use is already imposing huge costs too.

[Sir John Hayes]

I am interested in the right hon. Lady’s comment on productivity, because it is one of the prevailing macroeconomic challenges facing this Government and previous Governments, as the stalling of productivity has been a problem across western economies for a while. It is true that AI might help, but it could also harm. Equipping people with skills is critical to building productivity, and dependence on AI may actually undermine people’s confidence and their skills over time.

[Anneliese Dodds]

I absolutely agree. That phenomenon of so-called cognitive offloading and the lack of scaffolding in skills is a genuine concern. Equally, we see how rapidly and accurately AI can assess tumour scans, for example, working with human beings to deliver the highest possible quality of healthcare. I agree with my hon. Friend the Member for Camborne and Redruth (Perran Moon) about the potential there. We need to have a sensible approach to this issue.

I also want to mention some additional harms. We are now seeing ubiquitous slop on the internet, AI-enabled fraud and cyber-attacks, and the impact of AI summaries on local media use as well, so I want to see the CMA using its powers for competition-focused action on strategic market status services. I want the Government to adopt a maximalist approach to the “Watch this Space” White Paper, and I want us to stop treating bots as separate, because AI is increasingly going to be integral to our interface with the internet, as well as to so much else, so we need to adopt a safety-by-design approach to online regulation, and that needs to recognise the increasing ubiquity of AI.