[DigitalToday reporter Yoonseo Lee (이윤서)] Two out of three office workers have used artificial intelligence tools that are banned at their companies for actual work, a survey found.

Tech outlet TechRadar reported on June 18 that 66 percent of office respondents in a survey by digital operations management company PagerDuty said they used AI tools that were not allowed for work.

Unauthorized AI use appears to go beyond misconduct by a handful of employees. The share was higher at 72 percent at companies with 1,500 or more employees and reached 60 percent even at small and midsize firms. It suggests the practice of using AI outside internal policies is spreading regardless of company size.

The issue is less about use itself than concealment and the risk of data leakage. One-third of respondents who used AI for work said they would intentionally not inform a manager or superior that they were using it. Some 30 percent said they hide it because company rules are overly restrictive or they worry about colleagues' reactions, while 29 percent said unclear internal rules make them reluctant to disclose it.

The survey also found a lack of trust in internal AI policies. While 86 percent said their organisation has an AI policy, 81 percent perceived that executives are held to different standards. The view that executives receive exceptional treatment in AI-related decision-making and policy compliance was stronger at larger companies.

Employees rated their own AI capabilities higher than those of the policy function. Some 72 percent of all respondents said they understand how to use AI in their work better than the AI governance team does. That figure rose to 80 percent at companies with annual revenue of $1 billion or more, and senior executives held that perception more strongly than lower-level managers.

That confidence led to policy workarounds. Some 44 percent said they used AI tools to avoid constraints in company-approved work software, and 38 percent said they shared AI-assisted output without separate disclosure. Some employees also used AI tools or large language models on personal devices to avoid leaving traces.

The risk of leaking sensitive information was also significant. Some 43 percent admitted entering emails or work-related data into public AI systems. More than one-third said they put customer information into public AI systems, and 31 percent said they uploaded financial information, confidential company documents and even internal business strategies. Because such tools operate outside company systems, it is difficult to control how information is handled after submission.

Penalties after detection were inconsistent. When AI policy violations were found, more than half ended with informal guidance, but 48 percent received formal discipline. It showed organisations are failing to strike a balance between rapidly growing reliance on AI in the workplace and enforcement of rules.

The survey shows it is difficult to control workplace use simply by trying to block AI adoption itself. Employees are finding workarounds that prioritise productivity and convenience over rules, and unclear policies and perceptions of executive exceptions are amplifying that trend. As a result, companies' AI management tasks appear to be shifting beyond simple bans to deciding which tools to allow and to what extent, and how to prevent the input of sensitive information.