AI misuse among staff rivals phishing as security threat

Employee misuse of artificial intelligence is now responsible for more serious security incidents than phishing, according to research from IO.

The business resilience specialist's 2026 State of Information Security report, The State of AI Security, found that 25% of UK and US cybersecurity managers who had experienced an incident in the past 12 months attributed their most significant security incident or near-miss to employee misuse of AI. This compared with 24% who cited phishing.

Examples included giving AI tools unauthorised access to business systems and sharing confidential information through public generative AI platforms.

Using generative AI without permission or guidance was the most commonly reported employee security mistake, cited by 35% of respondents, compared with 32% who identified clicking suspicious links. Meanwhile, 22% said employees had incorporated AI into important workflows without the security team's knowledge and the same proportion reported staff sharing sensitive data with public AI tools.

Although 51% of organisations had updated their AI usage policies or guidance, the findings suggest that formal rules are not necessarily preventing unauthorised use.

Larger organisations were more likely to have formal, enforced AI policies than those with fewer than 100 employees, at 46% compared with 36%. However, they were also slightly more likely to report employees integrating AI into workflows without security team knowledge, at 24% compared with 21%.

Chris Newton-Smith, CEO of IO, said: "AI adoption has moved from the IT department into almost every corner of the organisation, but governance doesn’t appear to have made the same journey. The problem isn't that employees are using AI – it's that organisations don't always know which tools are being used, what they're being used for, what information is being put into them, or who is responsible for the consequences."

The research also identified uncertainty over accountability. While 34% of respondents considered the CEO ultimately responsible for business resilience, 27% said their organisation was shifting AI security and accountability from the security team to individual employees.

"This is why good governance of AI systems is so important. The answer is not to put the brakes on AI. It is to have the governance in place to use it with confidence," Newton-Smith said. “As technology and threats become more complex, organisations need to take a more structured approach – defining who is accountable, which tools can be used, what safeguards are required and how potential issues are identified and escalated."

He added that employee awareness needed to be supported by clear responsibilities and processes, rather than treated as a standalone solution. He also pointed to ISO 27701, ISO 27001 and ISO 42001 as frameworks for addressing data privacy, information security and AI management.



Share Story:

YOU MIGHT ALSO LIKE


Resilience Rooted in Reality
In this podcast, CIR speaks to CLDigital’s Tejas Katwala about why organisations must move beyond checklist compliance to build living, data driven resilience. He explains how rethinking governance, risk and compliance, breaking down silos and focusing on value streams can create sustainable, real time resilience that is rooted in the way businesses actually operate today.

Building cyber resilience in a complex threat landscape
Cyber threats are evolving faster than ever. This episode explores how organisations can strengthen defences, embed resilience, and navigate regulatory and human challenges in an increasingly complex digital environment.