The (alleged) moral imperative to use AI

14 Oct, 2026

I like it when people in power say the quiet part out loud. Last December my reflections on the TechUK Digital Ethics Summit noted that politicians appear to be endorsing the following line of thought:

1. The only way a Labour government can achieve its goals is by increasing spending.
2. The only way we can increase spending (given the so-called 'fiscal rules') is if there is economic growth.
3. The only way to achieve economic growth is to 'mainline AI into the veins'¹ of the country.

If you combine that with the belief that Labour's policy objectives are morally good and you infer from 'is good' to 'ought', then you get a moral requirement to push AI into everything whether it is wanted or not. That moralising tone about AI dominated the Starmer administration.

And now a senior figure in the administrative state has said it out loud. On 14 October 2026, the Financial Times reported an interview with Rachel Watson, the Director General of the Independent Office for Police Conduct. The point of the interview was for her to inform journalists, and thus the public, that she was investigating several police offers for misuse of AI, that the problem was probably more widespread, and that it could affect conviction rates. But she didn't want to send the message 'Don't use AI' because:

I think there’s a genuine moral imperative to embrace technology that clearly has potential benefits for productivity

Wow! I understand that she thinks her job involves, in part, holding the police to ethical standards, so she has a professional position to make moral pronouncements, but this goes one step further. She has moved from enforcing proscriptions to making prescriptions.

Analysis

Sometimes it is worth examining what someone says very carefully. This is a direct quotation in the FT article, so we can assume Watson used exactly these words. So let's look again:

“I think there’s a genuine moral imperative to embrace technology that clearly has potential benefits for productivity,” Watson said, adding that “sensible judgment and sensible policies” were also required.

Let's start with the addendum. While I have serious concerns about the 'You must use AI (but not in a bad way)' style of leadership in universities, it makes more sense as a position for the IOPC to take. After all, the police already have tools and powers which have the potential to do significant harm to citizens if misused. The IOPC is one of the safeguards we have in place as part of the social contract which allows the police these powers and it should be ensuring that new technologies are not misused.²

Next, let us consider the choice of words 'moral imperative'. An imperative is an order, and when A gives an order to B, A is claiming authority - a form of power - over B. Someone who talks of moral imperatives rather than duties, requirements or obligations, is giving morality a very distinctive authority.

To see this consider the fact that most moral philosophers think that there can be 'pro tanto' duties, requirements and obligations. A pro tanto duty is a duty which is not always and in every case the right thing to do. So we might think we have pro tanto duties to keep promises, but in a particular case keeping my promise might result in something far worse than breaking my promise, so it is right to break the promise, even if I still need to apologise. Plato imagined the situation of being asked to return someone's sword to them - as promised - when they turn up in a murderous rage.

There can be no pro tanto imperatives. There can be hypothetical imperatives, but there can be no pro tanto categorical imperatives. This is all in Kant and anyone using the language of moral imperatives is - consciously or not - invoking Kant. It would be impossible to make the maxim 'embrace technology' a Kantian Categorical Imperative.

However, to be fair to Watson, the maxim she enjoins us to follow is actually:

embrace technology that clearly has potential benefits for productivity,

I will deal with 'productivity' separately, so let's just focus on an imperative to do something which 'clearly has potential benefits'. The use of 'clearly' marks that Watson is making an empirical, and contingent, claim about the benefits of the technology. We could argue about that, but instead let's focus on the inference Watson is making from clear (potential) benefits to a moral imperative. How does that work?

There are many things we could 'embrace' which have very clear benefits, but it does not follow that we morally should do them or even that they are morally permissible. That is the point of Jonathan Swift's Modest Proposal. So at the very least, Watson is assuming that 'embracing the technology' is permissible.

Perhaps that is true for some uses of 'AI' by the police in the analysis of information or writing of documentation, though that might depend upon whether we think that using technologies built on and perpetuating injustice and colonialism which also destroy the environment is permissible. However, it is obviously contestable in other uses, such as predictive policing. If there is a moral imperative here, it is to realise that there may be morally impermissible uses of technology and avoid them.

Even if the means for achieving the 'clear' benefits is obviously benign, we have to face the problem created by that word 'potential'. It is not that we can't have a moral requirement to do something which only has a potential benefit, for we can be required to try to save someone's life, rather the problem is the nature of this qualification.

If someone does not have enough money to feed their children, so uses their last £2 to buy a lottery ticket, this 'clearly has potential benefit', but of course there is no moral requirement to do it. Thinking about whether this is a good idea, let alone a requirement, involves taking into account the chances of success, of achieving the benefit, and the opportunity costs of pursuing it.

So if these potential benefits of the police using AI are to ground a permission, let alone an imperative, we need to know, it needs to be clear, that:

The expected benefits (i.e. probability times value) are greater than the expected costs

The costs of 'embracing AI' are huge for money and time is diverted from existing policing to pay for the systems, train staff, etc. That is money that could have been spent on more police officers and support staff. This is inevitable: probability = 1.

No one yet has any vaguely reliable way to measure the benefits of using AI systems anywhere, let alone in policing. We already see lies and obfuscation from various police forces about the effectiveness of live facial recognition, and the companies that seek to profit from police contracts only offer best case scenarios or fantasy. And even if we could reliably quantify the benefits, they are not guaranteed and require continual effort and 'keeping pace' to sustain. The value of the benefits is not known and whatever it is, it is far from inevitable because it may not materialise for a variety of reasons. So the benefits have to be greater than the costs in proportion to the uncertainty of achieving those benefits.

Productivity

At this point Watson performs the moral magic trick and tells us that the benefits are for 'productivity' - this is meant to trump all the concerns I have so far expressed.

Before we get into a critique, we should just note the logical structure of the move Watson makes by replacing the word with another:

I think there’s a genuine moral imperative to embrace technology that clearly has potential benefits for the standards of spelling in police reports

The idea that such a benefit creates a moral imperative is clearly insane. In contrast, other substitutions have the opposite effect:

I think there’s a genuine moral imperative to embrace technology that clearly has potential benefits for preventing injustice

This shows that the way Watson is using the term 'productivity', she is expecting everyone to accept it is a moral good.

Obviously, I hope, productivity is not a moral good, for then the less productive process for producing some X would always be morally worse. If you need an example, and one which will appeal to the readers of the FT, one might note that the newspaper is full of advertisements for very expensive mechanical watches. By any measure of productivity, owning a mechanical watch is less productive than owning a cheap quartz or digital watch. The latter will be more accurate, more durable, need less maintenance etc. Does that make it morally better? Are the wealthy purchasers of mechanical watches who choose a less efficient means of time-keeping making a morally bad decision?³

Perhaps, then, we might think that Watson is only claiming that productivity in policing is morally good. Lurking behind that thought is probably the neoliberal dogma that productivity in public services reduces cost to the taxpayer and that is a moral good (because taxation is a necessary evil). This raises two questions:

• Will embracing AI actually reduce cost to the taxpayer?
• Might a less productive police force be (morally) preferable?

Here is an argument for the answers 'No' and 'Yes' respectively: productivity is a measure of output per person.

• Where a task completed by a human can be completed by a machine, you will always improve productivity by automation. It does not follow that automation is cheaper. There have been historical cases where it is and also where it is not. The jury is still out on AI, not least of all because we do not know the true cost of the highly subsidised tools being sold at the moment nor the business plan to monetise them later. But the tech sector has a long history of pursing a business model of getting customer lock-in then either raising the price or reducing the quality or both. See enshittification, passim.

• Humane policing is clearly better than inhumane policing: even if you think that does not apply to some groups of people you have demonised, it applies to how you are policed. The more people - humans - involved in policing the less likely it is to become inhumane, for two reasons. One is that those inclined to forget their humanity will remain the minority they are; the other that the more people the greater the chance of a whistleblower.

Concluding comment

Watson's remark shines a light on the dangerous groupthink among those who regard themselves as leaders in our society: politicians, senior civil servants, senior management of businesses and corporations.

We must not let this rubbish go unchallenged.