The King’s Speech last week gave us our first glimpse of the Government’s priorities for the new Parliament.
From a tech perspective, in the run-up to the King’s Speech, the press were widely reporting that an AI Bill would be included. Although artificial intelligence (AI) did get a mention, there are two bills sitting under the Department for Science, Innovation and Technology and not one of them has AI in the title. These are:
- Digital Information and Smart Data Bill: which our team has summarised here.
- Cyber Security and Resilience Bill: which, considering the severe impact of the IT outage across the private sector and critical national infrastructure last week, seems very well timed.
In the King’s Speech itself, the Government stated it will “seek to establish the appropriate legislation to place requirements on those working to develop the most powerful artificial intelligence models”.
What does this mean?
This ties back to the Labour manifesto which stated that a Labour Government would ensure the safe development and use of AI models by introducing “binding regulation” on companies developing the most powerful AI models, and with what Peter Kyle (the now Minister for Science, Innovation and Technology) said at London Tech Week in June 2024: “At the moment there’s a voluntary code regulating AI, particularly frontier AI […] We would legislate to require the frontier AI labs to release their safety data. That’s to make sure we legislate the standards that are already in the voluntary code”.
There is some discussion now on how the Government might achieve this and under which bill – perhaps by granting a Secretary of State the power to create secondary legislation in relation to codes and standards? This could possibly be as part of the Cyber Security and Resilience Bill as the cybersecurity aspects of LLMs is something the AI Safety Institute has continued to focus on and, following the July global IT outage, this will not abate. Watch this space.
However, what does look possible is that for the many solutions underpinned by the “most powerful models” the Government may be hoping (with some justification) that by putting these standards on a statutory footing this will both wash through all industries, build greater trust and increase adoption. As companies move to increased AI adoption with potentially increased investment in digital, it will be interesting to see the impact of increased workers’ rights and whether this leads companies to increasingly focus on AI governance. Especially as employee rights and cybersecurity obligations are strengthened across all sectors and industries under this new Government and two significant and well known risks of implementing AI solutions are the cybersecurity of the solution and whether outputs produce discriminatory effects. Of course this may already take place by virtue of compliance with the EU AI Act for those providers or users with exposure to the EU market or EU customers.
What is clear is that while an AI Bill is not imminent, the use or reliance on outputs of AI are now well publicised and likely to get increased scrutiny, so a “buyers beware” approach is necessary in the absence of an AI governance regime that assesses and mitigates the risks while ensuring AI forms part of a digital transformation journey with maximum use and efficiency gain.
