Canadian workplace safety regulator WorkSafeNB (New Brunswick) has warned employers that using artificial intelligence to draft safety documents does not transfer their legal responsibility to comply with New Brunswick’s occupational health and safety laws. The advisory was published on September 10, 2026.
Errors range from minor to serious
Consultant John Debly, whose team reviews this type of documentation, noted that employers are increasingly relying on AI-generated material without sufficient expertise to verify its accuracy. Identified errors ranged from incorrect regulatory citations and regulations mixed up between Canadian provinces, to outdated standards and conflated American and Canadian guidelines. In some cases, the AI text even contained fabricated sources, including invented journal articles and fictional books with fake ISBNs. Debly noted that AI-generated text often reads authoritatively regardless of accuracy, making it easy for employers to skip verification steps.
A recommended approach rather than a ban
WorkSafeNB is not calling for a ban on AI, but recommends that employers craft detailed prompts requesting references, section numbers and links, anchor prompts to New Brunswick’s legislation or the Canadian Centre for Occupational Health and Safety, have qualified personnel review drafts, and consult WorkSafeNB’s prevention team when in doubt before documents take effect. Debly stressed that using AI provides no legal defence and cannot satisfy due diligence requirements: employers remain responsible for the accuracy and suitability of their safety documentation. The warning fits within a broader, international discussion among regulators about the risks of unchecked AI use in drafting compliance documents.
Source: WorkSafeNB warns AI can’t replace due diligence in safety policies, September 17, 2026
