Health and Safety Round-Up: The Latest HSE Statistics, GB CLP Updates, and AI Rules Take Shape

The current picture in UK and European health and safety

If you are a SHEQ director preparing an audit file or a board update, the current regulatory picture hands you a mix you cannot ignore: HSE's latest injury statistics, a fresh chemicals amendment to absorb, building procedures moving through the calendar, and a maturing set of AI rules that reach directly into industrial safety technology. This round-up sets out the changes worth diarising and points to what belongs in your next committee agenda.

The statistics and the enforcement record

The 2024/25 annual release is the primary reference for your injury and ill-health reporting this cycle. Read the estimates directly from the GOV.UK release and the HSE statistics overview so the figures in your board paper are traceable. HSE's own statistics point at the same recurring failure modes year after year, and the prosecution record continues to reflect them.

Enforcement focus in construction and logistics stays where it has been for years: vehicle-pedestrian interactions and PPE compliance. These are not exotic hazards. They are the forklift in the aisle and the operative who steps into an exclusion zone on the hundredth hour of a shift, when attention thins and the banksman is looking the other way. The HSE Prosecution feed is a useful standing check on which of these failure modes are drawing enforcement action.

Regulatory calendar: what commences and what is coming

The regulatory picture is busy but tractable. Osborne Clarke's regulatory outlook and HSE's own pages set out the near-term changes worth diarising.

• •GB CLP: HSE has published changes to the GB CLP Regulation. If you handle or store classified substances, check your labelling and safety data sheet processes against the changes set out on the HSE GB CLP page.
• •Asbestos and other forthcoming rules: HSE's forthcoming regulations page lists the changes moving through the pipeline, including work on the Control of Asbestos Regulations. Track that page for confirmed dates before you commit them to a plan.
• •Higher-risk buildings: Osborne Clarke's outlook covers the building procedures affecting higher-risk buildings; confirm the applicable commencement dates for your jurisdiction against that source before relying on them.
• •Martyn's Law: guidance on the Terrorism (Protection of Premises) Act 2025 is relevant to any operator running publicly accessible premises. The Osborne Clarke outlook tracks its progress.
• •EU medical devices: Commission Implementing Decision (EU) 2026/193 revises harmonised standards under the Medical Devices Regulation. The specific standards affected are listed in the EUR-Lex decision and the LNE-GMED summary if your products fall in scope.

For most heavy-industry sites, the common thread is documentation. Whether the driver is CLP, higher-risk building procedures, or an ISO 45001 management system sitting alongside machinery risk assessment, regulators keep asking the same question: can you show the control was in place, and that it worked? ALARP and UKCA/CE conformity remain the frame everything else hangs on.

The EU AI Act now reaches into safety technology

This is the development that changes how safety technology gets specified. Under the EU AI Act, many industrial vision systems are classified as high-risk when they act as a safety component, which triggers risk management, data governance, human oversight and conformity obligations. The classification rules in Article 6 are the reference point, and the Dataiku summary sets out the high-risk requirements and the phased timelines; confirm the applicable dates against those sources before you build them into a specification.

In practice, this means a camera that only records for later review sits in a very different regulatory place from one that acts in the moment. HSE has been clear that existing health and safety law already applies to AI. Its regulatory approach page and the Industrial Safetytech Regulatory Sandbox both push specifiers towards dedicated AI risk assessments rather than trusting marketing claims.

Where active safety fits the hierarchy of controls

None of this replaces the banksman, the yellow line, the mirror, the training or the supervision. HSE's recurring failure modes make the opposite case: existing controls need reinforcing, not removing. Active computer-vision systems belong in the hierarchy of controls as a last line that catches the breach a tired human misses, not as a substitute for the people doing the work.

This is the design principle behind SAiFI. It is an active safety platform: it detects hazards such as vehicle-pedestrian conflicts and exclusion-zone breaches and triggers a physical intervention, such as a barrier, a gate, a projected warning or an alarm, in real time. It scales from one or two cameras in a single high-risk zone up to a multi-zone deployment, without tearing out what already works. The value is in configuration: zones, rules and schedules set to match how a site actually operates, rather than a generic model dropped on top of a real yard.

What to do with this before your next committee

• ✓Restate the injury trend for your board, citing the HSE 2024/25 annual release and the HSE statistics overview directly.
• ✓Cross-check your CLP labelling and safety data sheet processes against the changes on HSE's GB CLP page.
• ✓Track HSE's forthcoming regulations page and confirm commencement dates for higher-risk building procedures relevant to your jurisdiction.
• ✓Where you use or plan to use AI vision as a safety component, complete a dedicated AI risk assessment and confirm human-oversight readiness against the EU AI Act high-risk requirements.
• ✓For vehicle-pedestrian and exclusion-zone risk, review whether your current controls detect and act, or only record for review afterwards.