Invisible Risks: Employees Are Using AI—and Employers Aren’t Watching

A recent report titled “Trust, Attitudes and Use of Artificial Intelligence: A Global Study 2025,” released by KPMG and the University of Melbourne, reveals attitudes toward and use of AI gleaned from more than 48,000 individuals in 47 countries. Additional survey information provides specific insight into how employees in the United States are using AI at work, perceptions of their own AI literacy and AI’s impact on their work product.

Why it matters: In the early phases of AI integration, companies have focused on AI performance gains and establishing corporate governance policies. KPMG’s research emphasizes the need for ongoing, proactive, management of how individual employees and teams are using AI.

By the numbers: In the U.S., 70% of those surveyed say they expect AI to deliver on a range of benefits, but only 28% have had formal or informal training on how to use AI. Against this backdrop, the research highlights what it calls employees’ “complacent use” of AI:

• 44% used AI at work in inappropriate ways.

• 57% made mistakes in their work due to AI.

• 44% used AI in contravention of company policies and guidelines, 46% of whom admit to uploading sensitive company data on public AI platforms.

• 58% relied on AI output without evaluating its accuracy.

Much of how and whether employees use AI at work is unknown to managers, with 50% of employees presenting AI-generated content as their own. The gap between employee use of AI and company awareness of use is an important area of focus as AI becomes an integral part of work teams.

The report makes recommendations for leaders centered on ways to bridge this gap including:

• Invest in general AI literacy to help employees learn how to safely and effectively collaborate with AI.

• As part of creating a culture where employees feel comfortable discussing their AI use, create structures where management and employees can engage in open discussions.

• “Fully integrate training and guidance on responsible AI practices into everyday workflows – including onboarding processes, project work, and performance reviews” – to create workplace norms for AI.