April AI policy update: AI adoption surges into a transparency vacuum

UK AI adoption climbs, Stanford flags a transparency crisis among frontier firms, and the battle over AI’s economic fallout intensifies from Washington to Beijing.

By James Boyd-Wallis

UK updates

UK businesses accelerate AI adoption but skills gap bites

A new report from Amazon Web Services (AWS) reveals that nearly two-thirds of UK organisations now use AI, and 68 per cent of adopters report productivity gains.

However, nearly half of organisations cite AI and digital skills shortages as the main barrier to adoption. While the UK has gained momentum with world-class research, talent and a vibrant startup ecosystem, momentum alone is not enough, the report argues.

AWS puts forward three policy priorities: closing the digital skills gap through training, public-private partnerships and AI literacy; helping organisations move beyond basic adoption; and scaling AI across public services so the government leads by example.

DfT builds AI tool to analyse public consultations

The Department for Transport (DfT) has collaborated with Google and the Alan Turing Institute to build a consultation analysis tool to process citizen feedback from public consultations.

The department manages around 55 public consultations each year, which can generate more than 100,000 free-text responses, making manual analysis time-consuming.

The new tool uses Gemini models to identify and categorise themes. Initial evaluation suggests it has achieved 90 per cent accuracy and could save the department £4m annually.

However, an Alan Turing Institute report on the tool notes it was systematically less accurate at mapping themes to responses for specific demographic groups. To mitigate issues of bias, accuracy and fairness, the department is employing a human-in-the-loop approach.

But human-in-the-loop approaches are not infallible. Will human reviewers have enough time to keep up with the pace of analysis? And will they trust their own judgement over AI output?

Liberal Democrats pitch plan for AI digital independence

The Liberal Democrats set out their strategy for British digital sovereignty, pledging to award billions in public contracts to British tech firms.

The party argues that the UK’s dependence on foreign-owned AI and cloud infrastructure poses a direct threat to national security and democratic processes.

To overcome these risks, the plan focuses on “strategic sovereign public procurement”, ensuring that critical national functions are powered by technology the UK can control.

Other proposals include establishing an AI regulator to enforce a statutory code of ethics, where safety and bias testing are designed in, and a national online crime agency to combat the threat of AI scams.

While the Liberal Democrats can struggle to cut through, it will be interesting to see whether any other parties react to, or borrow from, any of their proposals.

Global updates

Stanford AI index: capabilities up, transparency down

Stanford’s 2026 AI index, published this month, reveals that AI capabilities are advancing quickly but transparency is worsening.

The index, which has tracked AI’s evolution since 2017, shows that frontier models now meet or exceed human capabilities in areas such as PhD-level scientific questions, multimodal reasoning and competitive mathematics.

However, the largest AI companies are also among the least transparent, increasingly keeping training code, dataset sizes and parameter counts to themselves.

In many ways, this is not surprising. Businesses developing the most capable models will want to keep such data proprietary to retain a competitive advantage. But for policymakers, it comes at a cost: they are less able to assess the risks posed by such models. The question is whether voluntary sharing agreements will continue to meet policymakers’ needs.

US cements pre-deployment testing with AI developers

The Center for AI Standards and Innovation (CAISI) at the US Department of Commerce’s National Institute of Standards and Technology has announced new agreements with leading AI developers to conduct pre-deployment evaluations.

Reflecting broader concerns around AI and national security, the evaluations and targeted research aim to better assess frontier AI capabilities and advance the state of AI security.

The agreements enable government evaluation of AI models before they are publicly available, as well as post-deployment assessment and other research. Kevin Klyman, who led some of this work while in the US government, has more in his post.

Warren warns AI failure could trigger financial crisis

US senator Elizabeth Warren has warned that an AI bubble could trigger the next financial crisis.

Speaking at a policy event in Washington DC, covered by the Verge, Warren said she believed the technology has “enormous potential” but warned that if AI companies cannot increase revenue quickly enough, they will not be able to service their debt levels. A big stumble could trigger destabilising losses in the financial industry, she claimed.

Warren wants riskier investments separated from less risky ones, and wants Congress to refuse to bail out the industry if it fails.

China rules firms cannot fire workers to replace them with AI

In China, a court has ruled that companies cannot terminate employees simply to replace them with AI systems. The decision comes as authorities aim to balance the domestic labour market with a global race to develop AI technologies.

The court decided that a tech firm in eastern China had illegally fired one of its workers after he refused to accept a demotion when his job was automated by AI. As AI continues to develop, these issues will undoubtedly become a legal battleground and attract policymaker attention.

New views

In conversation with Sophie England

Continuing the Appraise Network’s AI policy leaders series, we spoke to Sophie England, head of UK public policy, AI and wearables at Meta.

In the interview, Sophie discusses the strengths of the UK government’s approach to AI policy but cautions against a growing sense of inconsistency where parliament is not fully aligned with the government’s agenda.

She also discusses the potential of open source AI to boost the economy and help businesses innovate, and talks through new fellowships to help the government make greater use of open source models.

Chartered PR practitioner James Boyd-Wallis is MD of tech and AI focused corporate and public affairs agency Highbury Communications and co-founder of AI policy network, Appraise.