Respondents linked unchecked AI output to incorrect filings, missed deadlines and failed audits, and teams that put sensitive data into personal AI tools reported problems far more often.

Among compliance leaders surveyed by regulatory intelligence company Enhesa, teams that never verify AI-generated regulatory guidance were nearly three times as likely to report a regulatory fine as teams that do, 37% compared with 13%.

Forty percent said their teams always or often act on guidance produced by AI without an expert reviewing it. The most common reasons were pressure for speed and efficiency (47%), trust in the tool's accuracy (42%) and urgent regulatory deadlines (34%).

More than half of respondents, 56%, reported at least one negative outcome tied to AI output that was wrong or unverified. Over the past 12 months, those included incorrect filings (30%), regulatory fines (21%), missed deadlines (20%) and failed audits (18%). Teams that rarely or never skip expert review reported fines at 7%, compared with 27% of all other teams.

Seventy-six percent of teams use AI across their whole compliance function or let it run fully autonomously on at least one core task. Among the 37% using fully autonomous or agentic AI on at least one task, 71% reported a negative outcome, compared with 47% of the rest.

In the past year, 64% of teams entered sensitive material into personal or noncorporate AI tools. Those teams reported negative outcomes at a rate of 82%, compared with 9% of teams that entered no such material.

Most respondents said they want AI answers they can defend. Ninety-one percent said it is essential for AI output to cite its sources before they act on it, and 88% want a complete record of how the AI arrived at its answer. Only 32% said they were very confident their team could justify an AI-driven compliance decision.

Respondents also reported strain from regulatory change. Ninety percent cited at least one business impact in the past year, including higher legal or consulting costs (53%), much heavier workloads (41%) and delayed product launches or market entry (34%).

Asked what would best help them stay compliant over the next 12 months, 43% chose AI combined with expert human counsel, 39% chose AI tools alone and 4% chose expert counsel alone.

Enhesa surveyed 273 senior compliance, risk, EHS, legal and sustainability decision-makers online in August. All work at organizations with 5,000 or more employees operating in at least two countries. Enhesa sells an AI-based regulatory compliance platform covering EHS, product compliance, chemicals and sustainability.

• Friday New Product Roundup: October 16

  This week's products include cut-resistant apparel, carbon-toe work boots, pallet rack repair kits, an automated compressed air control and a weatherproof cable enclosure. Read Now

• Workplace Noise Can Affect More Than Your Hearing

  Long-term exposure to hazardous noise can affect worker safety, overall health and quality of life, making consistent hearing protection essential. Read Now

• Why Hands-On Training Still Matters in a Digital Age

  VR, AR and AI can expand training opportunities, but physical practice remains essential when workers must handle equipment, coordinate with teams and perform correctly in high-risk situations. Read Now

• The Heat Stress Standard and the Employer's Decision Dilemma

  With OSHA's proposed heat rule unfinished but enforcement expanding, the metric an employer chooses to trigger protective action — Heat Index or wet bulb globe temperature — may matter more than the regulation itself. Read Now

• Navigating Electrical Safety’s Gray Areas

  Electrical safety standards provide a framework, but real-world work doesn't always offer clear-cut choices. Five common gray areas show why understanding the hazard, the task and the worker’s qualifications matter when the right decision is not immediately obvious. Read Now

• The Keys to a Successful Smart PPE Rollout

  A rushed rollout produces resistance. A focused pilot and transparency on goals produce a program workers trust. Read Now