Bureaucracy-as-a-service: The grim, grey future of AI government

As public debt soars and growth sputters, Prime Minister Keir Starmer and his Labour government have hailed AI as the magic bullet that will save money across the public sector and speed up services ranging from planning to justice.

Fine words, for sure. But the truth is that the dawn of "computer says no" government risks delivering friction and impermeable bureaucracy to the residents of this sceptred isle - creating an opaque state instead of improving transparency.

The government has talked a good game on AI ever since it came into office. Last year, Chancellor Rachel Reeves claimed this tech would "grow our economy, make our public services more efficient and open up new opportunities to help improve living standards."

But just this week, the UK Government shocked the tech sector by announcing that phone manufacturers will be forced to censor nude images directly on devices.

In the worst-case scenario, the policy gives this or future governments the ability to censor messages at the point of production - so that bad words and banned concepts cannot even be written, let alone shared.

At best, it creates bloatware, which slows down phones across the realm and fails to tackle the problem it intended to. Let's not forget that the social media ban for under-16s in Australia has been a complete failure, with an estimated 60% of kids still very much able to access accounts.

In Britain, we will soon have to identify ourselves to do pretty much everything online. It started with asking people to prove they are 18 to access pornography - a policy which ended up locking down people's phones until they were able to prove their age. I was asked to show my iPhone a form of national ID - despite the fact this doesn't even exist properly in the UK yet.

If Westminster builds huge databases showing who's proved their age to apply for permission to send nudes and watch pornography, it's hard to see that anyone will benefit apart from hackers offered a treasure trove of highly sensitive data to steal and ransom.

So what else does the UK Government have in store for us?

We reviewed government tenders and official announcements to find out what AI projects Westminster is spending our money on right now - and speculate about what the future will hold in password-protected Britain, as well as the inevitable challenges any attempt to roll out AI across public services will encounter.

Justice is not just blind, but opaque

Justice is one of the most visible areas in which AI is soon to be deployed.

Deputy Prime Minister and Lord Chancellor David Lammy said it would be used in the justice system to speed up trials and clear backlogs. He previously announced plans to radically restrict the right to a jury trial - a move expected to yield only a "modest" impact.

Right now, AI is used to transcribe conversations with offenders using a tool called Justice Transcribe.

But the tech is already creeping into the legal decision-making process. Judges are already planning to use a new AI tool to help "identify trial-ready cases and group similar hearings together" to "maximise judicial, prosecutorial and court resources to resolve cases sooner and deliver swifter justice for victims".

Assuming the project is successful, it will make court proceedings faster and more effective. However, it also raises the spectre of fast, summary justice that removes human judgement from processes that clearly require it.

The rollout of AI is already troubling legal minds. Earlier this year, a senior UK judicial body warned lawyers against uploading client documents into large language models after identifying AI-generated hallucinations in a legal filing by an immigration solicitor.

Cops are already using AI to scan videos for evidence, and the London Mayor's Office will expand that capability starting later this year, spending almost £500,000 to build "video searching with AI" systems.

AI will also be used in asylum claims, with the government recently announcing plans to use it to assess the age of applicants - a move the British Association of Social Workers (BASW) described as "gambling" on tech that's "prone to mistakes".

Elsewhere, Westminster plans to use AI as a tutor as well as a bureaucrat, with ministers aiming to provide personalised AI tutoring to as many as 450,000 disadvantaged pupils a year. Ministers are also funding AI bootcamps and apprenticeship schemes for young jobseekers, while commissioning research into whether the technology is already contributing to a decline in entry-level hiring.

The voice of authority

Soon, AI will speak on behalf of the state.

The government has just signed a memorandum of understanding with ElevenLabs, which claims to offer the world's "most realistic" voice AI.

Meanwhile, the Department of Work and Pensions is building a "conversational AI platform" to steer callers to the right person and offer "personalised call deflection". In other words, there's a strong chance that soon you will phone the DWP to ask why it's cut your benefits and you'll be personally told why it can't help, pointed to the website for further information and then roundly ignored.

Which is likely to be an experience shared across government services thanks to a new initiative called GovVoice. Again, Westminster promises this will "enhance the overall citizen experience by enabling automated handling of routine enquiries"

"This approach is designed to reduce wait times, improve accessibility, and provide quicker resolutions for common queries, supporting a more responsive interface with government services," the government wrote.

READ MORE: The Silicon Age Collapse: Systemic risks that could derail digital civilisation

More than £10 million of taxpayers' money will be spent on GovVoice, which will be "applicable across the whole public sector landscape" and offer " consistent, interoperable voice capability that can operate securely within government environments and scale across multiple services".

Anyone who has dealt with a call centre knows what this is likely to mean: impenetrable barriers and unanswered demands. There are limits to the capabilities of any AI model, and you bump up against them every time you present it with a query outside those it's been programmed to handle.

When you face this impassable obstacle, you intone or bash out the words "speak to human". But what happens when there are no humans left to answer your call because the UK can't afford to pay real-life civil servants anymore? It's not such a far-fetched scenario.

How I stopped worrying and learned to love the bot

Now, despite the tone of this article, we note there are clear benefits to using AI in public services. Cutting down on busywork, automating low-skilled tasks, and streamlining access to public services around the clock are clearly noble goals.

Unfortunately, the employment of AI creates three risks that are unlikely to be satisfactorily mitigated, or even understood, by many politicians.

1) Process opacity: It's already difficult to get public bodies to share information. This will become even more difficult when services are delivered and administered by machines whose inner operations are a black box.

2) Tech concentration: We do not expect the UK public sector to suddenly start developing frontier models which compete with those produced by Anthropic, OpenAI or other Big AI firms. If all of government depends on a small number of providers, cybersecurity risk will soar and the likelihood of dependency cascades intensifies.

3) Debasement of human capital: Automate too heavily without investing in the employment of skilled humans and the government risks building systems that none of its workers understand properly. We've seen this knowledge gap issue across industries, from mainframe to nuclear. If the government is forced to cut the size of the public sector and turns to AI to fill the gap, there is a strong chance that it builds systems so byzantine that no one understands them anymore - a major risk vector.

READ MORE: European Central Bank summons lenders over AI security threat to banking infrastructure

The most immediate concern is not the imminent creation of a dystopia, but competence.

Building an AI-powered state is not like launching a website or buying a new software package. It requires extraordinary technical expertise, institutional discipline and an understanding of second and third-order effects that governments rarely demonstrate.

History is littered with public-sector IT projects that ran over budget, failed to deliver or created new problems while attempting to solve old ones.

The danger is not necessarily that Westminster uses AI to create a dystopia. It ends up creating a bureaucracy so complex, opaque and automated that nobody fully understands how it works anymore.

The promise is efficiency. The risk is a state that becomes less accountable, less transparent and less human at exactly the moment citizens need those qualities most.

Perhaps Westminster's AI revolution will deliver faster courts, better services and lower costs. We hope it does.

But if Britain’s leaders are wrong, the consequences will not be borne by ministers, consultants or technology vendors. They will be borne by the public, who will find themselves trapped inside systems they cannot understand, challenge or escape.

And unlike a real-life bureaucrat, an algorithm cannot be persuaded, reasoned with or held to account. But it can tell you no - and if there's no human to appeal to, you're in trouble.

Follow Machine on LinkedIn