The Government’s new AI Opportunities Action Plan promises much: £14 billion in investment, 13,250 new jobs, and a vision of Britain leading the global race in artificial intelligence. It sounds bold and hopeful. But like most grand government projects, the promises deserve a closer look. What lies behind these glowing ambitions is a mix of potential progress, risks to liberty, and the same old flaws of state inefficiency.

Efficiency or Control?

The Government assures us that AI will improve our lives. Teachers will use it to tailor lessons, doctors will diagnose illnesses faster, and bureaucrats will speed up planning applications. These claims are easy to cheer for—who wouldn’t want less red tape and better services? But efficiency in government often comes with a price: greater power to monitor and control its citizens.

Artificial intelligence thrives on data, and its application in governance means surveillance. AI can track movements, predict behaviours, and analyse patterns in ways humans never could. While it might help manage traffic or prevent crime, the same technology could just as easily be turned to monitoring protests or flagging dissidents. A government armed with AI is a government with unprecedented power to observe and interfere in daily life.

Without firm safeguards, the technology that promises convenience could lead to control. As Orwell himself might warn, efficiency is not always a virtue when wielded by those with unchecked authority.

AI’s Genuine Promise

It would be unfair to dismiss AI entirely. Like all tools, it depends on how it’s used. Even in a government context, AI could solve real problems. Take the NHS, for instance. Overburdened doctors spend countless hours on paperwork; AI could take on this administrative drudgery, leaving clinicians free to focus on patients. Faster diagnostics and better data sharing could save lives and money.

Consider, too, the misery of dealing with government call centres. The underpaid and overworked staff often provide slow, unhelpful, and occasionally biased responses. AI systems could replace these inefficiencies with quicker, impartial answers. Done properly, automation could strip out the frustrating human element from many interactions with the state.

Some might argue that the state should not exist at all, and these functions should be left to the private sector. That’s a fine theoretical position, but for now, the government is here to stay. If it must exist, we may as well demand it runs more efficiently.

The Hidden Cost: Energy

Here’s where the cracks in the plan become obvious. Training and running AI systems require colossal amounts of electricity. For example, training a model like OpenAI’s GPT-3 consumed about 1.287 gigawatt-hours of energy—the equivalent of powering 200 average British homes for a year. And that’s just the training phase. Running AI at scale, as envisioned in this plan, demands even more.

A single data centre dedicated to AI operations could consume between 100 and 200 megawatts annually, comparable to the energy use of a small city. Britain’s current energy policies, with their emphasis on wind and solar, cannot support this level of demand reliably. These renewable sources are intermittent by nature, leaving gaps that no amount of wishful thinking can fill.

The only practical solution is nuclear power. Unlike wind and solar, nuclear provides consistent, high-capacity energy with low emissions. If the Government is serious about AI, it must abandon the ideological rigidity of Net Zero policies and embrace a pragmatic approach to energy generation. Without this, the lofty promises of the AI Action Plan will collapse under the weight of their own power bills.

Governments and Inefficiency

Even with enough energy, AI’s success hinges on its implementation. Here, history gives little reason for optimism. Governments are not known for competence. Corruption and mismanagement plague almost every public sector project. Why should AI be any different?

Consider past IT initiatives. The NHS’s disastrous patient records system wasted billions without delivering results. Welfare systems frequently crash under the strain of poor design. AI, being far more complex, risks amplifying these failures. Contracts will go to the usual suspects—overpriced consultants and inefficient firms—while politicians boast of progress that exists only on paper.

Worse still, AI could be twisted to serve political agendas. Imagine systems trained to prioritise the party line or to subtly suppress dissenting views. Far from a tool for better governance, AI could become a weapon for entrenching power.

A Flicker of Hope

Yet, despite these risks, there is hope. AI can reduce costs and make legitimate functions of government less burdensome. Automating routine tasks could shrink bloated bureaucracies, freeing resources for meaningful work. Data-driven policymaking could replace ideology with evidence, assuming the data isn’t cherry-picked.

For these benefits to materialise, the Government must approach AI with humility and discipline. Transparency and accountability are essential. Without these, even the best technology will be squandered.

Conclusion

The AI Opportunities Action Plan is a grand vision, but it is also a gamble. It offers a chance to improve public services and boost the economy, but it comes with serious risks. AI’s energy demands clash with Net Zero policies, its power threatens civil liberties, and its implementation will likely be marred by the usual inefficiencies of government.

Artificial intelligence is a tool, not a saviour. It can make governments smarter, but it can also make them more dangerous. Whether it becomes a force for progress or oppression depends on how it is used. For now, we should watch carefully and demand that this powerful new tool is wielded with wisdom, not recklessness.