The Fire Service May Not Attend: AFA Attendance Policy in 2026
Three fire and rescue services have now decided that an alarm signal alone is not enough to justify a response to a commercial building. If your fire strategy assumes an alarm brings an appliance, in several parts of the country that assumption is now wrong.
INFIRISK Team·10 min read·
Shelly groisman, CC BY-SA 4.0, via Wikimedia Commons
On 1 July 2026, North Yorkshire Fire and Rescue Service stopped sending fire engines to automatic fire alarms at premises where nobody sleeps, unless a fire has actually been confirmed.
It is not an isolated decision. Devon and Somerset made a bigger change eight months earlier. London has been operating a version of it since 2024. Taken together, three services covering a substantial share of the English population have now decided that an alarm signal, on its own, is not enough to justify a blue light response to a commercial building.
If you are the responsible person for a workplace, that changes something fundamental about your fire strategy. The assumption that an alarm brings the fire service is no longer safe, and in several parts of the country it is now wrong.
What each service has actually done
The precision here matters, because none of these is a blanket withdrawal and all three are routinely described as one.
North Yorkshire, from 1 July 2026. The service says it is alerted to over 2,500 false alarms every year. Its published policy is that it will continue to attend all incidents where a fire has been confirmed, will not attend automatic fire alarms at premises where people do not sleep unless a fire has been confirmed, will attend alarms at identified high risk premises based on service intelligence and risk information, and will continue to attend all alarms at premises where people sleep and at domestic properties.
Devon and Somerset, from 4 November 2025. The changes went further. Industrial and commercial premises moved to 24 hour non attendance unless there are confirmed signs of fire. Retail and public assembly premises did the same, with out of term time cover maintained for schools. Residential premises are moving to daytime non attendance on a phased basis, with care homes, houses in multiple occupation and sheltered accommodation exempted, and the phasing starting with prisons, boarding schools and hospitals that are not already exempt.
The scale behind that decision is stark. In the five years to March 2025, automatic fire alarms alerted the service almost 58,000 times. On average, only 1.7 per cent of the alarms it attended were classed as emergencies. The service expects to save £178,577.53 a year.
Chief Fire Officer Gavin Ellis was careful about the distinction that most commentary blurs. "We will always attend if someone has reported a sign of fire. The changes are only to what we will do in response to an automatic alert, not a report of a fire."
That is the whole policy in two sentences. The service has not stopped responding to fires. It has stopped responding to signals.
Devon and Somerset also note that this extends a policy introduced 12 years ago, when the service stopped attending alarms at non domestic properties during weekday working hours without confirmation. The direction has been consistent for over a decade. What has changed is the pace.
London, daytime hours.London Fire Brigade's policy is that it does not attend automatic fire alarms between 7am and 8.30pm unless someone calls to confirm there is a fire. Eight categories of building are exempt and still get an attendance without confirmation: residential care homes, nursing homes and hospices; specialised housing including sheltered and extra care sheltered housing and supported living; student accommodation and halls of residence; hotels, motels, B&Bs and other guest accommodation; hostels including homeless, rehabilitation and youth hostels; prisons, young offender institutions and other secure establishments; Grade 1, Grade 2 and Grade 2S listed heritage buildings; and buildings of substantial public significance identified as exempt by the Brigade.
The Brigade's own figures explain why. In 2025, automatic fire alarms made up 34 per cent of all incidents it attended, some 47,500 calls. It states that less than 1 per cent of calls from non residential alarms are ultimately recorded as fires.
The national picture
The Ministry of Housing, Communities and Local Government published fire and rescue incident statistics for the year ending March 2026 on 22 July 2026. False alarms are now the single largest thing the fire and rescue service does.
There were 255,047 fire false alarms in England, up 1.6 per cent on the previous year's 250,987. Against a total of 641,227 incidents, that is 40 per cent of everything attended. Fires accounted for 27 per cent and non fire incidents for 34 per cent.
The composition is worth understanding, because the three categories behave very differently.
Due to apparatus, meaning the system itself signalled without a fire: 172,247, down 2.6 per cent.
Good intent, meaning a person genuinely believed there was a fire: 77,181, up 13 per cent.
Malicious: 5,619, down 2.1 per cent.
Apparatus false alarms are the ones a responsible person can actually do something about, and they are the ones these attendance policies target. They are also two thirds of the total and slowly falling. Good intent calls, which no maintenance regime will fix, rose sharply.
The charging power, and its limits
The three services above are also part of a wider pattern of charging, and this is an area where a great deal of loose figures circulate. Here is what is actually in the legislation.
Section 18C of the Fire and Rescue Services Act 2004, inserted by the Localism Act 2011 and in force since 1 April 2012, permits a charge for responding to a report of fire only where four conditions are all met. The report is of fire at premises that are not domestic premises. The report is false. The report is made as a direct or indirect result of warning equipment having malfunctioned or been misinstalled. And there is a persistent problem with false reports at the premises made as a direct or indirect result of warning equipment under common control having malfunctioned or been misinstalled.
Read condition (c) again. A staff member burning toast and pressing a call point is not chargeable. A contractor setting off a detector with dust is not chargeable under this section either, unless you can characterise it as equipment malfunctioning or being misinstalled. The power is narrower than most published summaries imply.
North Yorkshire's charging position statement is a useful worked example because it publishes the arithmetic. A charge is made for the fourth and subsequent calls received in a rolling 12 month period from premises to which the Fire Safety Order applies. Its 2024 scale gives £46.75 for a crew manager, £130.50 for three firefighters and £175.75 for a fire engine, each for one hour, plus administrative costs of £63.00 for a station manager and £30.25 for a grade 6 administrator. The total invoice is £446.25. Buildings within a site are treated individually by Unique Property Reference Number, invoices go to the responsible person, and if the count drops below four in a rolling year no charge is issued. Appeals must be made in writing within 28 days and are answered within 21 days.
So the exposure for a building generating six chargeable calls a year is not trivial, but nor is it the four figure per incident sum that occasionally gets quoted. Check your own service's published scale rather than assuming a national rate, because there isn't one.
What changed in the standard
BS 5839-1:2025, the code of practice for design, installation, commissioning and maintenance of fire detection and alarm systems in non domestic premises, was published on 30 April 2025 and supersedes BS 5839-1:2017. Anyone still specifying against the 2017 edition, with or without its amendments, is working from a superseded document.
The false alarm provisions were restructured rather than tightened. The Fire Industry Association's guide to the changes, published in June 2025, records that the trigger points themselves are unchanged from 2017. A preliminary investigation is indicated where a system exceeds four false alarms per 100 detectors per annum. An in depth investigation is indicated for systems with more than 40 detectors that exceed five per 100 detectors per annum.
What is new is a formal method for working the rate out. A new Annex F gives the false alarm rate as Afr equals 100 multiplied by Afn divided by DAF, where Afn is the number of false alarms and DAF is the number of detectors and alarm devices. Two annexes also moved, with the previous Annex E becoming Annex D and the previous Annex F becoming Annex H, which matters if your maintenance paperwork cites annexes by letter. The revision also places greater emphasis on multi sensor detectors, which are the single most effective technical answer to nuisance activations in dusty or steamy environments.
If your system has never had its false alarm rate calculated as a rate, rather than counted as a raw number, that is the gap. Four alarms in a building with 30 detectors is a problem. Four alarms in a building with 300 is not.
The national guidance is twelve years old
There is one more thing worth saying about where the sector has got to.
The National Fire Chiefs Council's guidance on unwanted fire signals, which its own page describes as having been formally launched on 17 June 2014 at the CFOA FireEx Conference, is still the current national document. It succeeded the CFOA Protocol of 2010, and its text still refers throughout to CFOA, the organisation that became the NFCC in 2017.
That is not a criticism of the content, much of which is sound. It is an observation about consistency. Individual services are now making significant and materially different attendance decisions, and the national framework they are all nominally working within predates every one of them. A responsible person with sites in three counties cannot assume the same response in each, and there is no national document that will tell them what to expect.
What to do about it
If you are responsible for non domestic premises:
Find out, specifically, what your local service will do. Not the national position, not what it did five years ago. Every service above publishes its policy. Read the one that covers each of your sites.
Work out who is going to make the 999 call, and be sure they will actually make it. Where attendance depends on confirmation, someone has to walk to the panel, identify the zone, go and look, and then dial. That is a named role with a named deputy, not an assumption.
Check whether you are on an exemption list. If you run student accommodation, a care home, a hotel or a listed building in London, you are almost certainly still getting an automatic attendance, and your procedure should say so rather than adding an unnecessary confirmation step.
Calculate your false alarm rate per 100 detectors, using the Annex F method, and see where it sits against the four and five thresholds.
Look at the cause of every activation, not just the count. If they cluster in one area, that is a detector siting or detector type problem and a multi sensor swap may fix it permanently.
Read the fire risk assessment against the new reality. A strategy that relies on a rapid attendance to a signal, in an area where the service has said it will not attend a signal, is now an assessment that is not suitable and sufficient.
The last point is the one that carries legal weight. Attendance policy is not something a responsible person controls, but it is a change in circumstances, and the duty to keep a fire risk assessment under review does not have an exception for changes you did not ask for.
