UK and European Health & Safety Round-up: Enforcement, First-Aid Standards and the AI Act Deadline

The season in UK and European health and safety

Recent months have handed SHEQ directors a lot to file. The HSE published its Regulation Action Plan for 2026-2027, a builders' merchant was fined £2.2 million after a conveyor crush killed a worker, and the EU AI Act reached the enforcement milestone that pulls computer-vision safety systems into scope. Add an updated first-aid standard and a record period for European product recalls, and there is enough here for a board paper and a toolbox talk. This round-up pulls the threads together and reads them from the perspective of someone answering to a regulator and a board.

HSE sets growth priorities and names the technology

The HSE has published its Regulation Action Plan: Growth Goals 2026 to 2027. The document is notable less for any single rule than for its direction of travel. It sets out support for innovative technologies, including vision-based systems, an agile chemicals regime under UK REACH, and enablement of Net Zero transitions. For a regulator that has historically led with duties and inspections, the explicit encouragement of technology in the control hierarchy is worth reading closely.

This sits alongside the HSE's earlier mapping work on artificial intelligence, which frames vision-based active safety within existing ISO 12100 machinery-safety principles and the UK's ALARP standard. The regulator is not treating AI as a category apart. It is treating a camera that detects a hazard and triggers a barrier as part of the same control hierarchy that already governs guarding, isolation and safe systems of work. That is the right frame. It also raises the bar: any system placed in that hierarchy has to demonstrate it reduces risk, with documented human oversight and performance monitoring.

The £2.2 million reminder of what static guarding misses

On 27 March 2026 the HSE secured a £2.2 million fine against a major builders' merchant after a worker was killed in a conveyor crush. Enforcement of this kind turns on the same failure mode again and again: a hazard that was foreseeable, and a control that did not intervene in the moment it mattered.

That is the failure mode static controls are weakest against. A physical guard is either fitted or it is not, and once a route around it becomes routine, nothing intervenes until an investigation. This is the gap between reactive safety, which produces evidence after the event, and active safety, which acts in the moment. A vision system watching that conveyor line can enforce an exclusion zone on the hundredth hour of a shift, when a banksman has looked away and a worker has taken the shortcut they have taken safely fifty times before.

The March fine was not isolated. The HSE's press releases through 2026 record a steady run of prosecutions for the same recurring themes: machinery guarding, working at height and conveyor systems. The pattern matters more than any single case.

• 604,000

  self-reported non-fatal workplace injuries in 2024/25 (HSE)

• 124

  workers killed in work-related accidents, 2024/25 (HSE provisional)

• £2.2m

  fine after a worker was killed in a conveyor crush (March 2026)

The fatality figure is the one that draws board attention. It runs against the assumption that workplace safety improves steadily year on year, and it lands hardest in the sectors this audience runs: construction, manufacturing and warehousing.

The AI Act clock has moved

From 2 August 2026, further EU AI Act obligations have become applicable, including rules for general-purpose AI and the wider governance and penalty framework, with high-risk system obligations phasing in on the Act's staggered timeline. Computer-vision safety platforms sit within this scope. Any UK operator placing such a system on the EU market needs to prepare conformity assessments benchmarked against the Act, and to align with ISO 12100 and, in the UK, ALARP.

For SHEQ leadership the practical reading is this: compliance is now something to engineer in, not bolt on. Privacy-by-design under GDPR, UKCA and CE marking, documented human oversight and audit trails are no longer optional extras for a vision system in a regulated environment. They are the entry ticket. The upside is that a system built to that standard produces exactly the audit-ready evidence a board and a regulator want to see.

First-aid provision and a record recall period

Two more items belong in the file. The BSI has issued BS 8599-1:2019+A1:2026, the amended standard for workplace first-aid kit contents and the provision of AEDs, updating specifications with explicit attention to burns dressings, eyewash and defibrillator placement, all driven by site-specific risk assessment. Compliance with the standard is not mandatory where a risk assessment shows other provision is adequate, but the HSE continues to reference the BS 8599 series as a benchmark. If your first-aid provision has not been reviewed against the amendment, that is a short, cheap job worth closing out.

On the product side, European recalls hit an 11-year high in the first half of 2025, with events rising 10.2 per cent to 7,729. Food and beverage featured prominently among the causes. For regulated manufacturers, particularly in pharma, beverage and chemical, the recall data is a reminder that process deviation carries the same weight as physical hazard when the regulator comes to call.

The through-line for site leadership

Read together, these developments point the same way. The regulator is encouraging technology in the control hierarchy, enforcement continues to punish the gap between a known risk and an intervention that never came, and the AI Act now sets a clear conformity bar for the vision systems that close that gap. None of this replaces the banksman, the yellow line or the guard. It backs them up. The value of an active system is that it catches what a tired, distracted or unlucky human inevitably misses, and it acts before the incident lands rather than recording it for the investigation.

That is the case for active safety in a sentence. SAIFI Edge sits alongside existing controls and enforces exclusion zones, PPE compliance and machinery safeguards in real time, driving physical interventions such as barriers, gates and alarms in the moment a hazard appears. If any of this news maps onto a risk you already know you have, that is the conversation worth having.