One in Five UK Workers Hand Passwords to AI Tools, Survey Finds
A new report from credential security firm Dashlane suggests many UK employees are trusting AI tools with far more access than their employers realise. The report, titled Misplaced Trust: Inside the Workplace's New AI Security Blind Spot, draws on a wider survey of 3,800 people across the UK, US and France, including 1,000 UK workers and 250 UK IT executives.
The findings show 19% of UK employees have shared a password, API key or login credential with an AI tool or agent. Nearly half (48%) have allowed AI to carry out at least one activity without asking permission first, such as accessing emails, company files and customer information. A quarter (25%) said they use AI tools their employer does not know about.
The risks are significant. Information given to AI tools may end up on third-party servers that could be compromised, or be used to train models and later appear in answers given to other users. AI agents are also a possible target for hijacking by threat actors. Among IT and security executives, 72% said their organisation had experienced a security incident involving an AI agent's credentials or access in the past 12 months.
Yet 92% of those executives said they were confident they know about every AI agent operating in their organisation, which sits awkwardly alongside what employees report. Meanwhile, 68% of UK IT security leaders feel caught between pressure to deploy AI quickly and the need to stay secure. Dashlane's chief product officer, Christophe Frenet, argued that the basics of identity security, including just-in-time provisioning, least privilege, good credential hygiene and authentication, must now apply to AI agents as well as people.
Why it matters
The survey shows a gap between what security leaders believe they can see and what staff are doing: 92% of executives feel sure they know every AI agent, while a quarter of employees use tools their employer does not know about. Any business relying on an approved-tools list as its main control is exposed, because credentials pasted into an unmanaged agent bypass it and make misuse hard to detect or investigate. It forces a decision to manage AI agents as identities with their own limited access, rather than letting them borrow staff logins.
What to do
- Tell staff clearly that passwords, API keys and login credentials must never be entered into AI tools or agents.
- Review which AI tools can access email, company files and customer data, and remove permissions that are not needed.
- Ask employees which AI tools they use, so unknown shadow AI is brought into view.
- Apply least privilege and time-limited access to any AI agent you approve.