Stop work authority (SWA) is every worker's stated right and duty to halt any task they believe is unsafe, immediately and without fear of punishment, until the hazard is fixed and a qualified person clears the job to resume.
That sentence is easy to print on a wall and hard to make real. A stop work authority program lives or dies on one question: what happens to the person who actually pulls the stop? If the answer is a thank-you and a fix, you have a program. If the answer is a raised eyebrow and a slower shift, you have a poster. This post walks through the trigger, the stop-notify-correct-resume cycle, and the leadership follow-through that separates the two. It is educational, not legal advice.
What is stop work authority?
Stop work authority is a formal policy that gives every worker and contractor on a site the authority and the responsibility to stop a job the moment they see an imminent hazard, with no requirement to ask permission first. The key word is every. A new hire on day one has the same right to call a stop as the plant manager, and the policy has to say so in writing. SWA turns the vague hope that "someone would speak up" into a named, expected action that carries the backing of leadership.
It is not the same as a slow-down, a work refusal grievance, or a supervisor's shutdown order. Those are all narrower. SWA is broad and immediate: anyone, any time, for any condition they honestly believe is dangerous. It sits alongside your job safety analysis and your near-miss reporting system as one of the few tools that acts before the injury rather than documenting it after.
When should a worker use stop work authority?
A worker should stop work whenever a reasonable person in their position would believe the task carries a real risk of serious harm to a person or the environment, and there is not time to fix it through the normal chain. The trigger is honest belief, not certainty. SWA is designed to be used on the "I'm not sure this is safe" feeling, because that feeling is usually correct and usually early.
Common triggers look like this: a guard is missing or bypassed; equipment is energized during what should be a lockout; a load is rigged wrong; a fall hazard is open; the job has drifted from the written procedure; conditions changed and nobody re-checked the plan; or the simplest one of all, a person feels rushed into a shortcut. The cost of a wrong stop is a few minutes of lost production and a short conversation to confirm the job is fine. The cost of a wrong "keep going" can be a life, a lost hand, or a fire. That asymmetry is the entire argument for erring toward the stop, and it is why a good program never asks a worker to be certain before they act. A stop that turns out to be unnecessary is still a win, because it proves the authority is usable.
How does the stop-notify-correct-resume process work?
The process is short on purpose, so a rattled worker can run it from memory. Each step has one job.
- Stop the task and warn the area. Halt the specific job and tell everyone in the line of fire, using plain, direct words. There is no form to fill out first and no supervisor to ask. The stop comes before the paperwork, always.
- Notify the supervisor and affected crew. Say what you saw, where, and why you stopped, calmly and factually. The goal is a shared understanding of the hazard, not an argument about whether the stop was justified.
- Correct the hazard, or contain it. Fix the condition, or isolate and barricade it so no one can get to it, then have a qualified person confirm the fix actually removed the danger rather than hiding it. If the fix is bigger than the shift, the job stays down.
- Resume only on qualified clearance. Work restarts when someone competent to judge the hazard, not just the person under schedule pressure, agrees it is safe and says so. Restarting to "see if it's fine now" is not clearance.
- Log it and learn from it. Capture the stop, the cause, and the fix, then feed repeated triggers into design and procedure changes. A stop that keeps happening on the same job is telling you the job itself is broken, not the worker.
Notice that the human steps, 1 and 2, are the fast ones, and the judgment steps, 3 through 5, are where competence and honesty matter. Most failed SWA programs are not failing at step 1. They are failing at step 4, where schedule pressure pushes a restart before the hazard is truly gone, or at step 5, where nobody ever closes the loop and the same stop happens next week.
What makes stop work authority credible instead of a poster?
Credibility comes from what leaders do in the ten minutes after a stop, not from the policy document. Workers read the real rules off behavior. If the first stop of the quarter earns a public thank-you and a fixed hazard, the program is alive. If it earns a sigh, a "was that really necessary," or a quiet mark against the person, the program is dead and every future stop dies with it.
Three leadership behaviors make it real. First, respond fast and treat the stop as a gift of information, because the worker just handed you a hazard for free. Second, never let a stop cost the person anything, not pay, not standing, not their next assignment; this is the non-negotiable one. Third, close the loop out loud, telling the floor what was found and fixed, so the next person sees that stopping changes something. This is the same trust that underpins behavior-based safety observations and a strong safety culture: people report and act only when it is safe to do so.
What does the law say about refusing dangerous work?
Stop work authority is a company policy, but it rests on a legal floor. Under the OSH Act, workers already have a limited federal right to refuse imminently dangerous work, and it is illegal for an employer to retaliate against someone for exercising it. A good SWA program is more generous than the legal minimum, because it does not wait for the hazard to reach the "imminent danger" bar before a worker can act.
| Aspect | Federal right to refuse (the legal floor) | Stop work authority (a stronger company policy) |
|---|---|---|
| Trigger | Reasonable belief of imminent danger of death or serious harm | Honest belief a task is unsafe, well before "imminent" |
| Conditions | Asked employer to fix it, no time for normal enforcement | No preconditions; stop first, sort it out after |
| Scope | Refusing your own dangerous task | Stopping any unsafe work, including others' and contractors' |
| Protection | Anti-retaliation under the law | Written no-reprisal promise, backed by leader behavior |
What do the numbers say?
Context for why an early stop matters, from primary sources:
- The Bureau of Labor Statistics recorded 5,283 fatal work injuries in 2023 the large majority of them preceded by a hazard that was visible before anyone was hurt.
- OSHA affirms that workers have the right to refuse dangerous work and that retaliation for exercising safety rights is prohibited under Section 11(c); complaints of reprisal generally must be filed with OSHA within 30 days.
- OSHA's Recommended Practices for Safety and Health Programs put worker participation and management leadership at the center, the two ingredients an SWA program cannot function without.
Every one of those fatalities is a moment where an earlier stop was possible. SWA exists to move the decision from after the injury to the instant the hazard is still just a hazard.
Where stop work authority quietly breaks down
The failure mode is rarely a worker too timid to stop. It is the loop that never closes: a stop gets called, the immediate hazard gets patched, and the record of what happened lives in someone's head or a paper log that no one reads. The same trigger returns next month because the underlying condition, a bin out of reach, a procedure that no longer matches the machine, a rushed changeover, was never designed out. Harmony is an AI-native layer that connects machines, software, and paperwork into one operational layer, with no rip-and-replace, so a stop-work event is captured at the station as structured data, not a note on a clipboard. AI search returns cited answers across those records, so a supervisor asking about a job sees every prior stop on it, and Harmony's digital workflows route the corrective action to the person who can close it and confirm it stayed closed. It is not a safety-compliance product; it keeps the lesson from a stop from evaporating. The same paper-to-digital move Harmony makes on the production floor is described in the CLS case study.