As enterprises navigate their AI strategies, employees are using their own AI tools under the table.
A report by Wakefield Research and PagerDuty published Thursday found that, of 1,250 office professionals surveyed, two-thirds reported using AI tools at work despite lacking explicit permission from their organization to do so. This phenomenon, known as shadow AI, has plagued workplaces since OpenAI's initial launch of ChatGPT, and can create significant data security vulnerabilities.
According to the research, 88% of those surveyed have shared work-related information with public chatbots. Around 43% shared emails or other correspondence, 40% have shared meeting notes and summaries, 34% have shared customer data, and 31% have shared financial information or confidential documents.
Of those who have used shadow AI at work, 53% have received feedback to discontinue use, and 48% faced formal disciplinary action as a result.
The survey points to several reasons for this undercover AI use:
- Around 39% said they would prefer to use shadow AI rather than being told they can't use it at all. Meanwhile, 33% are avoiding scrutiny from management or leadership, and 30% are hiding it to avoid restrictive company policies and judgment from colleagues.
- While 86% of organizations surveyed have AI policies in place, 81% believe that their leadership operates under a different set of AI rules than lower-level employees, perceiving a double standard.
- Additionally, 72% of those surveyed are confident they understand AI better than their own tech teams do.
These driving factors, coupled with the fact that employees are wantonly putting information into public models, create a "massive enterprise liability," Tim Armandpour, CTO of PagerDuty, said in a statement.
"We know the demand for AI is there because we see it in our own platform," Armandpour said. "The goal for any executive today should not be to slow down AI adoption, but to redirect that energy into proven platforms that offer governance and automation at scale."
Our Deeper View #
AI has been improving at an exponential rate in recent months, as OpenAI, Anthropic and Google continue to leapfrog one another in capabilities. Though their models can do incredible things, they also present serious security risks when used in the wrong contexts. Additionally, because these models are changing so rapidly, it's becoming increasingly difficult for organizations and systems to keep up. It's why shadow AI — and employees' confidence that they fully understand a technology that even experts are struggling to grasp — can be so dangerous. Without more effective educational programs and a clearer picture of the consequences of using these models, employees are exposing their organizations to significant risk.