The PDF Does Not Make the Stunt Safe

A form is not a safety system

The suspicion that risk assessments are written for insurers is not irrational. Productions may be asked for them by studios, insurers, completion guarantors, locations, permit authorities or clients. A dated document also creates a record after the event. But public evidence does not support the stronger claim that one universal form is always an insurance condition, or that a particular page count automatically protects coverage.

The operational value sits in a chain of actions: identify the hazard; describe a credible failure path; choose controls; resource them; assign them; brief the people affected; verify the controls in the real location; and reassess when an assumption changes. Break that chain and the PDF becomes evidence that paperwork happened—not that risk was managed.

There is no controlled, film-industry study showing that a standalone Stunt Risk Assessment form reduces a particular percentage of injuries. A broader 2007 systematic review found insufficient evidence to recommend for or against specific occupational health-and-safety management-system interventions. That uncertainty matters. It prevents a form from borrowing scientific authority it has not earned.

It does not follow that hazard assessment is useless. Official investigations repeatedly identify unrecognised failure modes, missing controls and unreviewed changes as contributing factors. The defensible conclusion is narrower: assessment can be useful because it makes those decisions explicit, but documentation alone is not the active ingredient.

No single accident invented the Stunt Risk Assessment

The familiar daily PDF has no single inventor. Three histories converged: general occupational-safety law began demanding systematic assessment; stunt work became a specialised, coordinated production discipline; and studios turned those duties into repeatable documents that could travel with scripts, call sheets and safety memos.

France already required a prior study plus protection and rescue measures for unusual filming risks in 1971. The European Union's 1989 framework directive then pushed risk evaluation and prevention principles into national systems. Germany's 1996 Arbeitsschutzgesetz made assessment and documentation explicit, while its production-specific rules already treated dangerous scenic action as work requiring protection, rehearsal and suitable people.

The forms became more visible from the 1990s onward because the legal process, specialist role and digital production workflow finally met. The 2010s and 2020s brought more public templates, initial-versus-residual scoring, revision fields and specialist plans. That is a history of standardisation—not proof that every production uses the system equally well.

The duty travels; the form does not

Germany, Britain and the European Union state the assessment duty more explicitly than US federal law, but none of those systems prescribes one global Stunt Risk Assessment layout. What must exist is the process: identify, evaluate, control, document, communicate and review.

In Germany, §§ 5–6 ArbSchG require the employer to assess work-related hazards and retain appropriate records. § 3 BetrSichV adds a competent, change-sensitive assessment before work equipment is used. DGUV Information 215-315 then translates the general duty into special scenic action and says the assessment's extent should follow the degree of hazard.

Britain requires a “suitable and sufficient” assessment. HSE's film guidance says production complexity should shape the management system and that an employer cannot delegate away the legal duty. The Stunt Coordinator may supply the technical truth; production remains responsible for making the system possible.

At US federal level, the General Duty Clause addresses recognised serious hazards, and OSHA publishes Job Hazard Analysis guidance, but there is no nationwide law requiring every stunt to use a document with that exact title, a particular matrix or a 24-hour deadline.

California's current pilot is narrower and more explicit. From July 2025 through June 2030, productions in Film and Television Tax Credit Program 4.0 must use a dedicated Safety Advisor and written general or specific assessments. Major stunts are among the activities requiring a Specific Risk Assessment. The rule does not apply to every California shoot, and the programme's own evaluation is not due until 2029.

Where common sense ends

“Use common sense” sounds attractive because nobody wants a document that warns experienced crew that gravity exists. HSE itself says assessment should focus on risks capable of real harm, not every trivial or fanciful possibility. But common sense is not shared technical knowledge, and it is not a control for information one department has withheld from another.

Common sense can reasonably cover conditions that are visible, stable, familiar and within an ordinary worker's experience—after the production has provided a safe workplace and the information needed to recognise them. It stops carrying the load when a hazard is hidden, specialist, moving, changed, compounded or outside the person's authority.

A camera operator cannot infer a modified vehicle path. Background cannot intuit a new exclusion zone. A performer cannot inspect an engineering assumption they were never shown. A new crew member may not recognise a local cue. Fatigue, language, darkness and time pressure can also move an apparently ordinary risk beyond ordinary judgement. The same interface problem appears in our map of Stunt Department roles and reporting lines.

How long is long enough?

There is no correct number of pages. A one-page daily assessment can be adequate for a tightly bounded, low-complexity action. Twenty pages can still be inadequate if they repeat generic policy while hiding the one changed assumption that matters.

In a private working archive reviewed for this article, the formats ranged from one-page German daily sheets to a 19-page general location assessment, beat-by-beat packs with initial and residual ratings, and narrative safety breakdowns. Length did not predict usefulness. Specificity, ownership, revision and a visible connection to the planned action did.

A useful layered structure solves the reading problem. The full assessment and method remain available to the people making and verifying decisions. Specialist annexes hold engineering, aviation, vehicle, effects or rescue detail. The crew-facing layer distils only what changes behaviour that day: scope, affected zones, responsibilities, cues, stop conditions and material changes.

The document is too short when another competent person—absent from the original meeting—cannot understand the activity, credible failure, people exposed, controls, owners and change triggers. It is too long when the critical instruction cannot be found, generic text obscures the scene-specific decision or the affected crew receive a technical archive instead of usable information.

Safe Work Australia makes the same cross-industry point bluntly: a safety method statement should be short, specific and easily understood; excessive detail can make it difficult to apply, monitor or review. The principle transfers well to film even though the cited document governs Australian high-risk construction work, not stunts.

What a plausible assessment contains

The assessment should describe decisions, not perform expertise. It does not replace the engineering plan, aviation procedure, effects plan, medical protocol or Stunt Coordinator's method. It connects those systems and makes their assumptions visible—the same reason fight previs must become a production decision tool, not merely an impressive clip.

Start with the exact creative and production scope: scene, activity, place, date, version and the people or departments involved. Then write hazards as events. OSHA's useful structure is environment, exposure, trigger, consequence and contributing factors. “Vehicle” is a category. “A changed path exposes a crew position after loss of control” is a scenario a production can act on.

Record the risk before additional controls so the document does not disguise the problem it was built to solve. Then move through the control hierarchy: eliminate or redesign first; isolate or engineer next; use administrative controls and PPE where necessary. A matrix can help prioritise. It cannot certify that a stunt is safe.

Every material control needs an owner, a time, a verification point and enough authority to be enforced. Residual risk is the decision after those controls—not a number chosen to make the cell green. The document must also state what change invalidates the assessment and who has authority to stop, revise and re-brief.

Risk Assessment, Method Statement and call sheet answer different questions. The assessment asks what can fail and what must control it. The method describes how the agreed work is organised. The call sheet communicates the day. They may travel together or appear in one RAMS pack, but one cannot silently substitute for the others.

What accident reports can—and cannot—prove

Accident reports are unusually clear about how risk-assessment systems fail. They do not prove that a better PDF, by itself, would certainly have prevented the event. They show which foreseeable questions were absent, disconnected from controls or not revisited when the work changed.

The BBC's report into Richard Hammond's 2006 Top Gear jet-car crash found that a supplier assessment was generic, contained the wrong location, treated all risks as tolerable and offered sparse mitigation for foreseeable failures. It also described a disconnect between controls written down and arrangements to implement and check them. Distribution was not the same as a suitable system.

WorkSafeBC's investigation into the 2017 Deadpool 2 motorcycle-stunt fatality found no adequate risk assessment addressing safety controls, speed and equipment limitations. A second transition ramp was added during rehearsal, yet the changed geometry and possible acceleration path were not adequately assessed; barriers were absent. The report also identified orientation, supervision and headgear failures. The lesson is not “add more rows.” It is “treat rehearsal information and physical changes as new evidence.”

HSE's Fast & Furious 9 case presents the third pattern: an assessment existed but omitted critical failure modes and between-take verification, while crash protection did not follow changes to the set and sequence. The £800,000 fine came after the production company pleaded guilty to breaching its statutory duty to people not in its employment.

Can a risk assessment be credited with preventing an accident? Sometimes a documented redesign, cancellation or added control provides a credible prevention story. But the strongest outcome—nothing happened—rarely produces a public investigation, and the avoided counterfactual cannot be measured directly. California's pilot is designed to collect final safety evaluations and is due to be assessed by 2029. Until then, claims of a quantified stunt-RA effect would outrun the evidence.

The final test: can it change the plan?

A credible Stunt Risk Assessment should be able to change at least one real production variable: the shot, action, exposure, people, equipment, place, time, resources, safeguards or decision to proceed. If nothing in the document can affect anything on the day, it is a record—not a control.

That is where common sense belongs too. Common sense is the last metre of human judgement inside a well-designed system. It is not the missing briefing, the hidden technical assumption or the budget line the production chose not to fund.

The PDF matters when it carries specialist knowledge across departmental boundaries and survives revision. It fails when it becomes a talisman: attached, signed, scored low and forgotten. The stunt is not made safer by the weight of the file. It is made safer by the decisions the file can still force.