Well, it’s been quite a week for AI in the public sector.
First, West Midlands Police admitted a controversial decision to ban visiting supporters of an Israeli football team was in part based on an AI hallucination.
Second, a map published to show railway investment in the northern of England bore no relation to geography and thought to be riven with AI hallucination.
As we know, a hallucination is an AI result which has been entirely made-up by AI.
No doubt, there will be an investigation where lessons to learn are available.
All this leads me to a timely analysis of my AI public sector tracker survey which I published last week in the Essential Public Sector Comms Playbook 2026.
It’s timely to go through that data and reach some conclusions.
The public sector is using AI but not always fessing up to it
In the survey I carried out in Autumn 2025, 80 per cent of public sector communicators say they are using AI either daily or weekly. Yet, little more than half of this figure – 44 per cent – work where there is an organisation-wide policy. Not only that, one in 10 have a comms team AI policy.
All this represents a stunning safety gap.
Using AI without guard rails means you can produce content that is flawed. Trust is in short supply at the moment and the sure way of burning through a chunk of it is to do something that people are not aware or in support of. Like, for example, using AI without being transparent.
With an AI policy you can build in safeguards and oversight.
Without it, it’s a free-for-all.
If I was ahead of comms in the public sector I’d by lying awake at night about this.
Not only that, but organisations are concealing how they use AI. In the survey, 45 per cent don’t admit to using it. Less than 10 per cent of public sector comms mark each piece of content that has involved AI in the creation.
All this is understandable. Telling people you are using AI may lead to criticism and awkward questions. A good media relations advisor would say that it’s wiser to choose the time and place to have that conversation rather than have it uncovered through an FOI request or some such like.
AI in the public sector is about ideas not images or video
It’s clear that the public sector is remarkably consistent in the way it uses AI.
Using AI for idea generation is used by 84 per cent of people. That’s an increase of eight per cent on the previous survey three months earlier.
Using AI for press releases is the second most single popular use with 39 per cent. That’s a rise of two per cent compared to the previous poll.
Image generation is the third most popular use with 20 per cent a decline of three per cent. Webpage creation (13 per cent) is fourth highest with editing video fifth (nine per cent). Creating audio accounts for just four per cent. That’s a fall of two petr cent.
The survey was carried out in Autumn 2025 with more than 320 people taking part.
What to do
The one thing not to do is ban the use of AI outright as a matter of principle. But I would be holding fire somewhat until you had a policy in place. This needs to be a set of principles on how this will be used.
As a first step, a Blue Peter knowledge of AI is needed by everyone in the comms team as a minimum. Beyond that, a policy that helps shape your approach would be essential.
For more, I deliver training to help you make sense of the changing landscape.
ESSENTIAL AI FOR PUBLIC SECTOR COMMS
ESSENTIAL COMMS SKILLS BOOSTER
ESSENTIAL VIDEO SKILLS REBOOTED.
Picture credit: Calculator By Piotr433 – Own work, CC0.