Everyone Signed Something

October 28, 2026

On 9 November 2016, at Carlsberg's brewery in Northampton, contractors were removing a compressor from a refrigeration system when ammonia escaped in a large, uncontrolled release. One contractor, 45, was killed. Another, 57, was seriously injured. Twenty other people needed hospital checks. It took several days to bring the gas down to a safe level.

The isolation on that compressor was a single valve, not proved shut. HSE's own guidance calls that the lowest standard available. What the job needed was positive isolation — the highest standard, the kind that doesn't rely on one valve staying where somebody left it.

Both men worked for Speedrite NE Ltd, a sub-contractor. Speedrite had been brought in by Crowley Carbon UK Ltd, the principal contractor Carlsberg had appointed to run the project. Carlsberg was the client. Three organisations, one job, one valve that should have been two.

In 2022, Carlsberg — by then trading as Carlsberg Marston's Brewing Company — was fined £3 million with £90,000 costs, pleading guilty to breaching sections 2(1) and 3(1) of the Health and Safety at Work Act. HSE's principal inspector didn't let the client off the hook for having hired someone else to manage the job: "Both the client, Carlsberg, and the Principal Contractor should have worked together to ensure that the risk was adequately managed. Not only Carlsberg had a duty here."

Four times the risk, every single time

HSE's own research found something that's easy to state and easy to forget: a worker is roughly four times more likely to be injured in their first month with a new employer than someone who's been in the same job for a year or more. New to a place, you don't yet know where the hazards actually live — not the ones written in the induction pack, the ones everyone on site has learned to work around without thinking about it.

A contractor is structurally in that first month every time, on every site, however many years they've been in the trade. Speedrite's engineers were presumably competent at removing compressors — that's the job they're hired for, over and over. What they weren't, and couldn't have been expected to be on their own, was intimately familiar with the specific isolation history of this specific refrigeration system at this specific brewery. That knowledge belongs to the host. Whether it reaches the people doing the work is a separate question, with a separate, much weaker set of guarantees behind it.

Everyone signed something

Nobody at Carlsberg, Crowley Carbon or Speedrite set out to leave a compressor isolated by one valve instead of two. Every one of those organisations almost certainly had a document somewhere that said the job had been assessed, the risk understood, the isolation arranged. That's the uncomfortable part. A paper trail existed. It just didn't amount to anyone, specifically, having checked that the isolation in place matched the hazard in front of them.

That's what a contracting chain quietly does to responsibility. Inside one company, a hazard like this usually has a traceable owner — a supervisor, a competent person, someone whose job it literally is to hold the whole picture in their head. The moment the work crosses an organisational boundary, each party has a reasonable basis for assuming the adjacent one has covered the part nearest to them. A client brings in a principal contractor on the reasonable assumption that competence to run the job safely comes bundled with the contract. That principal contractor, in turn, is relying on the client's own site team to have flagged whatever's dangerous about this particular plant — the client knows it, after all, and they don't. Further down, a sub-contractor was hired to do one job well, not to coordinate the wider picture, so naturally assumes somebody above them already has. Each of these is a sensible thing to believe. None of it gets checked against what the others actually did.

Four years before Carlsberg, the same gap showed up at Chevron's Pembroke refinery in Wales in a different shape. On 2 June 2011, five contractors from two different firms — B&A Contracts and Hertel — were emptying a tank in the Amine Recovery Unit with a vacuum tanker, what the court was told was meant to be a routine job. An explosion killed all four of them. HSE's investigation found the risk posed by flammable atmospheres in that unit "were not understood or controlled," on top of what it called longstanding failures in the refinery's own safety management systems. Two experienced contracting firms working side by side, each competent at their own trade, and the hazard that killed four people sat in the gap between what either of them had been told. Valero, which had taken over the refinery by the time of sentencing, was fined £5 million with £1 million costs; B&A Contracts was fined alongside them.

The industries are different and the gap is the same. Nobody was lying on the forms in either case — the forms just aren't built to catch the one thing that falls between two people who each had good reason to think the other had it.

What the law actually asks

This isn't a gap the law missed. The Management of Health and Safety at Work Regulations 1999 have two provisions written for exactly this situation. Regulation 11 requires employers sharing a workplace to co-operate and co-ordinate with each other. Regulation 12 goes further: a host employer must give any outside employer working in their undertaking comprehensible information on the risks to health and safety that the host's own operations create, and on the measures in place to deal with them. Carlsberg and Chevron were prosecuted under the broader duties in sections 2(1) and 3(1) of the Health and Safety at Work Act — to their own employees, and to people not in their employment — but regulations 11 and 12 are the specific clauses that name the problem this piece is about: one employer's hazard knowledge, and another employer's people doing the work.

Read regulation 12 closely, though, and the gap is still there, just moved one level down. It requires "comprehensible information." It doesn't require that the information actually land with the person opening the valve, or that anyone check it has. A document can satisfy the regulation and still fail the job — which is more or less what a single, non-proved valve dressed up as an isolation record manages to do. The law sets the floor for what must be communicated. It was never going to be able to legislate for whether anyone on either side actually understood it.

It diffuses afterwards, too

There's a detail in the Carlsberg case worth sitting with. HSE prosecuted Crowley Carbon, the principal contractor, as well as Carlsberg. But by the time the case reached court, Crowley Carbon had been placed into compulsory administration by its creditors. There was no longer a company there to prosecute.

That's not a loophole anyone engineered. It's just what happens when a contracting chain's accountability runs through a company that may or may not still exist by the time anyone's called to account. If you're the client, hiring a principal contractor to take on much of the day-to-day responsibility for a job doesn't make your own liability shrink to match what's left of theirs. Carlsberg's fine was the same size whether or not Crowley Carbon was still standing to share it. Your organisation is usually the one that's still there.

Five things to check tomorrow

Pick a job currently running on site that involves a contractor and find the single person, by name, on each side who has actually walked the hazard through together — not whose signature is on the RAMS, but who had the conversation.
Find a job with more than one contracting layer on it — a principal contractor and at least one sub-contractor — and trace exactly where the site-specific hazard information is supposed to cross that boundary. A form passed down isn't the same thing as a briefing given.
Check the isolation standard specified for a recent or current contracted job against what the hazard actually needs, rather than what's routinely used for that kind of task. A single valve reads as "isolated" on paper the same way positive isolation does.
Ask what happens to your oversight of a job if your principal contractor stopped trading tomorrow. Who, on your side, actually knows what they knew?
Find two contractors working in the same area at the same time and ask who, specifically, is coordinating between them. If the honest answer is nobody, that's the gap this piece is about.

The question worth asking

A brewery and a refinery, five years apart, with nothing in common except the shape of the failure: competent people doing a job they understood, inside a structure that never quite integrated what each of them knew. Everyone involved could point to something they'd signed; what nobody could point to was a moment where the whole picture actually got checked against itself.

That's not a reason to trust contractors less. B&A Contracts and Hertel were specialists; Speedrite's engineers knew compressors. The gap wasn't in anyone's competence at their own trade. It was in whether that competence ever met the host's own knowledge of the hazard, in the same room, at the same time, held by someone who could act on both.

So: on the next job that crosses an organisational boundary on your site — client to contractor, or contractor to sub-contractor — who is the one person who has actually seen the whole picture, rather than just their own slice of the paperwork?