![](https://emergencyservicestimes.com/wp-content/uploads/2025/05/ESS-Button-resized.png")
Spend enough time in ICT and you will encounter a familiar pattern. A new concept emerges, proliferates, attracts training and standards, generates products claiming to use it, then either beds in or fades out.
Artificial Intelligence (AI) has existed for decades but has only in recent years become mainstream, appearing across everyday tools from word processing to mobile phones, and generating its own ecosystem of qualifications and AI-branded products. Some of these are sophisticated, some little more than legacy chatbots. So far, so familiar. Yet AI has subverted the ICT lifecycle.
AI has impacted the social and workplace arenas profoundly. It can fix formulae, turn natural language into code, generate photorealistic faces or simply tell a joke. While the social sphere is doing what it normally does – impatiently pushing the limits – the emergency services sector does not have this luxury; it must adopt new technology safely and at a controllable rate, not something easily done with AI.
AI exerts pressure on its users: adopt quickly or be left behind, or, worse, become vulnerable to adversaries using the technology.
For such a user-friendly concept, AI is remarkably difficult to ring-fence. It permeates software, blurs the boundaries of personal and working life and, crucially, is evolving at pace. If we are Achilles, AI is the tortoise; every time we reach where it was, it has moved on. We don’t yet know the limits of AI, yet those limits continue to expand.
The MFRS approach
Prior to 2024, MFRS managed AI as an emerging but ambiguous technology, to be monitored and managed on a case-by-case basis. It quickly became apparent that this was not a viable approach. The evolutionary speed of AI meant that a long-term, sustainable strategy was needed.
To this end, in 2025, MFRS introduced its first substantial AI measure: the AI Service Instruction. This contains a number of key elements designed to protect the organisation and its users, while aligning our use of AI with the wider MFRS purpose: Here to serve, here to protect, here to keep you safe.
The most difficult aspect, though, was defining the concept. While it is accepted that AI is used in a host of applications and products in the emergency services, our focus was on accommodating the prevailing level of AI understanding, as well as describing the predominant AI tool our staff use: AI assistants. The definition was informed by looking at established definitions, such as those from the BBC, IBM and the ICO. It is a dynamic definition. Currently, our definition is:
‘AI is the demonstration of intelligence by a computer system, including dynamic learning, adaptation, apparent independent reasoning, and human-like behaviours. It enables technology to undertake a range of new and existing tasks with varying levels of human involvement.’
The broad scope of the definition reflects the reach of AI. Independent AI reasoning and completing tasks without human intervention are not aspects MFRS supports; human involvement is non-negotiable, but the ideas are included in order to illustrate to our audience – our staff – what AI appears to do now and what it may soon be capable of.
The Service Instruction continues with the common benefits and hazards, most of which the reader will be familiar with. In addition to the volume processing, speed and cost benefits, we recognise cyber security opportunities, advanced simulations (such as dynamic emergency calls) and plain language coding. Conversely, together with greater hacking and phishing sophistication, we recognise hazards such as the shorter – some would say non-existent – window between vulnerability discovery and exploitation, social upheaval, AI assistants becoming unstable and the deterioration of human abilities.
These factors inform the User Requirements, such as declaring when AI has been used in content creation, confirming all outputs, forbidding deepfakes and retaining social and academic skills. We explicitly prohibit the use of shadow AI, and within this we include entering personal or business information into AI models that are not approved by MFRS.
MFRS only allows personal and business data processing with Copilot in the M365 container. Because Copilot does not exist in a single form, we have a dedicated area on our intranet where the gateway to the ‘safe’ Copilot may be found. This includes an infographic background image that further reinforces AI benefits, hazards and user expectations.
This approach was hastened because mainstream applications have added, sua sponte, web-facing AI assistants to their programs. This is a practical illustration of the theoretical pressure discussed earlier – AI is emerging in everyday applications faster than organisations can assess the risks.
Training, partnership and professional bodies
MFRS does not exist in isolation; it works closely with blue light partners and professional bodies. The National Fire Chiefs Council (NFCC) has published a host of guidance on AI, including the AI Cyber Security Action Plan and the AI Ethical Framework.
Moreover, MFRS recently released a bespoke AI training module on its internal training system. It is a more interactive and user-friendly version of the Service Instruction, yet we are open to supplementing or replacing it with the upcoming NFCC AI module.
Senior support and AI champions
Support for any change needs to come from the top. We made a conscious effort to present AI to the Strategic Leadership Team (SLT) in tandem with our recommended approach. From these presentations, the next stage of the AI strategy emerged.
Currently, we are rolling out a trial of Copilot AI agents for selected representatives nominated by our SLT. The intention is for each key area (e.g. Finance, Legal Services) to have an AI champion who will use their agent to supplement their capabilities and feed back their experiences.
The future
The future ultimately depends on the results of the Copilot trials with our champions. It is hoped that agents may be used by all staff to support the work of MFRS teams. AI is not going away, but nor must it be rolled out at MFRS before we are ready.
MFRS has made significant progress over the past 24 months and the future will see AI become further integrated into the organisation. It is therefore imperative that we are prepared for iterations of AI that are currently known unknowns.
To read similar articles, check out our Technology channel.
Hear more from Mark at The Emergency Tech Show
Mark will be speaking at The Emergency Tech Show on 17 September (11:40-12:20) on the Emergency Tech Leaders’ Stage, sponsored by AWS. In the NFCC-led session, Major incident exercising and cyber-attacks: are you prepared?, he’ll explore how fire and rescue services are preparing for cyber threats as part of major incident planning. Join us at the NEC Birmingham on 16-17 September to be part of the discussion, register here.
