Newsletters
Risk Central
Digital EditionWorkers Say A.I. Helps Write Safety Documents, But Few Know Who’s Accountable for Errors
More than half of workers who have visibility into how their company creates safety documents say generative A.I. is already helping write them, yet just 18% report a formal policy spelling out who is accountable when that A.I.-drafted material contains an error, according to a survey of 514 U.S. workers by Tradesafe, a provider of industrial safety equipment.
The survey, covering workers in manufacturing, construction, warehousing, logistics and the skilled trades, found that 51% of those with insight into document creation say their organization uses generative A.I. for safety-related documents. One in five workers (20%) said they have personally witnessed an adverse outcome tied to an A.I. safety-document error, including near misses (11%), minor injuries (7%), failed inspections (6%), equipment or property damage (5%) and serious injuries (3%).
Oversight Gaps Widen Between Leadership and the Floor
Tradesafe found a sharp divide in how different levels of an organization perceive A.I. oversight. Only 49% of workers said A.I.-generated safety documents are independently reviewed before use, but the answer varied widely by role: 74% of company leadership reported independent review, compared with 52% of supervisors or managers, 36% of frontline workers and just 23% of safety professionals.
Frontline workers were also 45 percentage points less likely than leadership to say they have visibility into how safety documents are created at all, 31% vs. 76%.
More than a third of workers (34%) said A.I. adoption in workplace safety is outpacing their organization’s ability to properly oversee it, and 36% said they have personally encountered an A.I.-generated safety document with an outdated requirement, vague instruction or missing hazard step. Nearly half of workers (47%) believe an uncaught A.I. error could happen at their own organization.
Trust in these documents dropped once workers learned A.I. was involved. Tradesafe found that 63% of workers would trust a safety policy less knowing it was partially A.I.-written, a skepticism that ran higher among frontline workers (67%) than managers (61%). Conversely, 78% of workers said a qualified human reviewer increases their trust in a safety document regardless of whether A.I. helped write it.
When Documents Go Unchecked, Blame Shifts to the Top
Tradesafe tested how accountability shifts depending on how a flawed safety procedure was created, presenting respondents with one of three scenarios: a procedure written entirely by a person, an A.I. draft that was reviewed by a person, or an A.I. draft that went straight to the floor with no review.
When the unreviewed A.I. scenario led to a serious injury, 56% of workers blamed company leadership or the safety department (28% each), compared with just 9% who blamed the employee who handled the document. By contrast, when the mistake was in a human-written procedure, 49% blamed leadership or the safety department and 17% blamed the employee, and when the A.I. draft had been reviewed by a person, 39% blamed leadership or the safety department while 19% blamed the employee.
The pattern held when mistakes were caught before anyone was hurt: blame on the employee fell from 30% for the person-written procedure to 24% for the reviewed A.I. draft and 14% for the unreviewed one. Across all scenarios, an injury pushed responsibility further up the organization, with blame on company leadership nearly doubling from 13% when an error was caught in time to 24% when it caused an injury, while the employee’s share fell from 23% to 15% over that same shift.
Tradesafe also found that 64% of workers agree that blaming A.I. for a mistake usually signals the company’s own oversight isn’t clear enough. Asked what they would do if an A.I. instruction conflicted with what they knew was safe, 48% said they would ask a supervisor and 37% said they would stop and challenge it, while only 4% said they would follow it as written.
Herbert Post, vice president of Tradesafe, said the survey results raise questions for employers.
“What stood out is where workers place responsibility when an unreviewed AI-drafted safety document leads to an injury in the survey scenario. They look primarily to company leadership and the safety department. That raises an important question for employers: as A.I. becomes part of safety documentation, are review and approval responsibilities clear to the people creating and using those documents?” Post said.
Read more about the survey here. &
