A new survey from PagerDuty, an AI operations platform, has found that a large share of office employees are using AI tools like ChatGPT at work despite company policies that expressly forbid it, and many know exactly what they’re doing.
The study surveyed more than 1,200 office professionals across the United States, United Kingdom, Australia and Japan, all working at non-tech corporations or non-profits that generate or manage at least $500 million.
The key findings were:
- 72% of respondents said they believe they understand AI and its potential uses better than the managers who write their company’s AI policies.
- Among employees who use AI at work, 66% do so knowing it violates an outright ban on such tools.
- 48% of that group admitted they had already faced formal consequences for the behavior, yet continued anyway.
- 72% said they specifically go against their employer’s instructions regarding certain AI platforms.
- Rule-breaking was more common at larger firms (those with 1,500 or more employees).
The study also found that 45% of respondents use AI tools to help make work-related decisions, essentially outsourcing judgment calls to the technology. Separately, 43% admitted to pasting emails or other internal correspondence into public AI tools such as ChatGPT, Claude, or Gemini, that sit entirely outside their company’s own systems. That can mean sensitive customer data, financial records, or trade secrets ending up in tools never built to hold them securely.
The concern is that these platforms are not as private as users tend to assume. OpenAI’s own policy states that it retains user interaction data by default, unless someone actively opts out or uses a business-tier product like ChatGPT Enterprise. Even shared conversations carry risk: links created through a chatbot’s “share” feature are often publicly accessible and can be indexed by search engines, exposing the full exchange to anyone who finds the link, not just the intended recipient.
The findings suggest that blanket bans on workplace AI use are largely being ignored rather than obeyed.
PagerDuty’s research recommended that, instead of an outright ban on AI tools, companies should permit AI use within clear limits, invest in privacy-protected business versions of these tools, and maintain open conversations with staff about proper use, which is likely to be safer than relying on prohibition alone.
