When enforcement stops being clear cut

For a long time, enforcement has been based on clear rules. You were either speeding, or you weren’t. The line was visible, and the decision was relatively straightforward.

That’s starting to shift.

In Sussex, police are trialling AI-powered cameras that can detect drivers using their phones. Not just whether a phone is present, but how it’s being used, whether it’s in someone’s hand, and potentially even how it’s being held.

It’s a small change, but an important one. The system isn’t just detecting a rule being broken. It’s interpreting behaviour in real time.

Once you notice it, it’s hard to miss

And once you start looking for that shift, you see it elsewhere.

In north-east England, transport teams are using an AI “traffic twin” to model the road network and adjust traffic lights in real time. It predicts congestion and intervenes continuously, rather than reacting after the fact.

Within the Met Police, an AI tool has been used to analyse internal data on officers, surfacing behaviour more quickly. What might once have been occasional investigations becomes something closer to ongoing detection.

And in facial recognition systems, the shift goes further still. The system isn’t just identifying a face, it’s assigning an identity. When it gets that wrong, the consequences can move beyond error into investigation.

Across all of these, the pattern is the same. The system is doing more than observing. It’s shaping how behaviour is understood.

The line isn’t as clear anymore

This changes the nature of the decision. Speeding is measurable, a number, a limit, a clear breach. Phone use is less precise. It depends on context, position, and intent, and the same action might be interpreted differently depending on how it appears.

That doesn’t make enforcement unnecessary. Few people would argue that drivers should be using their phones. But it does change what’s being judged.

Part of the appeal is clear. These systems can operate at a scale that people can’t, processing more data, spotting patterns more quickly, and acting in real time. Things that might have been missed or delayed become visible almost immediately.

But that visibility can also create a sense of certainty. If something is flagged by a system, it can start to feel like a decision has already been made, even when the underlying judgement is based on patterns rather than context.

So who’s actually deciding

That raises a more difficult question. If a system interprets behaviour and that interpretation leads to action, who is responsible for the outcome?

Is it the organisation deploying the system, the people acting on its output, or the system itself?

In practice, responsibility tends to sit across all of these. But as systems move closer to real-time judgement, that boundary becomes harder to see.

None of this means these systems shouldn’t be used. In many cases, they improve outcomes, increase efficiency, and support organisations in ways that weren’t previously possible.

But the shift isn’t just about efficiency. It’s about moving from enforcing clear rules to interpreting behaviour, from reacting to events to continuously assessing them.

That changes the nature of what’s being decided. The question is no longer just whether a rule has been broken, but how much judgement we’re comfortable handing over in the first place.