Tony Blair’s institute is calling for governments to build a serious independent testing and certification industry for AI, rather than responding to public anxiety by either freezing adoption or waving it through. The proposal matters because it puts accountability, liability and practical checks at the centre of the AI debate, where slogans have recently been breeding rather freely.
Watch Desk analysis
What happened
A Tony Blair Institute report, published on Monday and reported by City AM, argues that governments should make AI assurance a national priority. That means independent testing, evaluation and certification of systems before and after deployment, alongside clearer rules about who is liable when they fail.
The institute says 38% of UK adults see a lack of trust as the biggest barrier to using AI, while 61% of organisations say reliability concerns have delayed or reduced planned investment. It estimates the global AI assurance market could reach $276bn (£204bn) by 2030.
The report also proposes using bounded, auditable and revocable AI systems inside government to bring together information, model policy consequences, track commitments and initiate pre-approved workflows. Political decisions, it says, should remain with people.
Why it matters
This is more than a call for another layer of paperwork. If AI is moving into public services, business operations and consequential decisions, “trust us” is a weak operating system. Independent testing could give buyers and citizens a way to compare systems, while liability rules would make the cost of failure less conveniently theoretical.
The proposal also turns assurance into an economic contest. Countries that develop credible testing expertise and standards early could sell those services internationally. The catch is obvious enough to deserve a bright marker: an assurance industry is useful only if its independence is real, its methods are inspectable and its certificates mean more than a logo on a sales deck.
Our read
Blair’s institute is right to reject the false choice between slamming the brakes on AI and letting it run wild. But “grip” needs a mechanism. Governments should start by specifying which systems require external testing, what evidence must be published, who can challenge a certification and what happens when a certified system causes harm.
The useful takeaway is practical: when an organisation announces an AI system, ask who tested it, what was tested, whether the results are public and who pays when it goes wrong. The future may be automated, but accountability still needs an address.
What to watch
- Whether the UK government adopts a formal AI assurance strategy.
- Which AI systems would require independent testing before deployment.
- Whether ministers set clear liability rules for failures.
- Whether assurance providers publish methods and results rather than issuing polished certificates into the mist.
Discussion spark: Should governments require independent certification for high-impact AI systems, even if the cost and delay make some deployments unviable?
Sources and evidence
- Tony Blair warns leaders to resist growing AI ‘techlash’ – City AM (21 September 2026, 11:44 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.