'Using AI does not replace your duty to exercise due diligence,' says regulator listing common errors
New Brunswick's occupational health and safety regulator is warning employers that artificial intelligence tools are producing flawed health and safety documents — and using AI to draft those policies does not reduce an employer's legal obligation to get them right.
WorkSafeNB published an advisory identifying a pattern of problems it has observed in AI-generated workplace safety documents. Among the errors are:
- policies that reference legislation from other Canadian provinces rather than New Brunswick
- documents that cite U.S. occupational health and safety standards
- content that is inaccurate or incomplete
- filings that omit workplace-specific hazards or required procedures
- documents that create privacy and confidentiality concerns, according to WorkSafeNB.
The advisory does not prohibit the use of AI in drafting health and safety documents, but it is direct about where accountability lies.
"As an employer, you are responsible for ensuring your health and safety policies, procedures and codes of practice comply with New Brunswick's Occupational Health and Safety Act and regulations and reflect your particular workplace hazards," WorkSafeNB stated.
AI doesn't know your workplace
The regulator noted that AI cannot replace an employer's knowledge of their own workplace — and that no AI tool can be expected to account for the specific conditions, hazards and regulatory requirements that apply to a particular operation in New Brunswick.
WorkSafeNB was explicit that no amount of AI assistance changes an employer's fundamental obligation.
"If you use AI to help draft health and safety documents, always review and verify the content before putting it into practice," the regulator stated. "Check that references to legislation, standards and regulatory requirements are accurate, current and specific to New Brunswick. Using AI does not replace your duty to exercise due diligence," according to the WorkSafeNB advisory.
AI errors play out across Canada
The WorkSafeNB warning arrives as AI errors are surfacing across a range of professional and institutional settings in Canada — with consequences that range from flawed patient records to fabricated legal citations, according to earlier reporting from Canadian HR Reporter:
- Ontario's auditor general released a special report in May finding that nine of 20 AI transcription systems evaluated for use by doctors fabricated clinical information, such as referrals and test orders that had never occurred.
- In a January 2026 Alberta human rights decision, a self-represented complainant cited a court case that did not exist, with respondent's counsel alleging it was the product of an AI tool. Tribunal member Rabie Ahmed confirmed there was "no such authority" as the case cited, and used the ruling to remind all parties — represented or not — to be aware of court notices governing AI use in proceedings.
A June 2025 analysis published by the Toronto Lawyers Association, authored by Cara Parisien and R. Scott MacKendrick of ROBIC LLP, documented the Federal Court, Tribunals Ontario, and the Trademarks Opposition Board all issuing practice directions requiring disclosure when AI has been used to generate filed content.
In British Columbia, a lawyer was ordered to personally pay a portion of opposing costs after submitting ChatGPT-generated fictitious case citations, they said. A similar incident before the Ontario Superior Court of Justice in 2025 resulted in a show cause order before the matter was resolved through an apology and a commitment to ethics training.
