Nearly one in five (19%) UK employees have shared a password, API key or login credential with an AI tool or agent, exposing businesses to greater security risk by giving artificial intelligence access to sensitive corporate information and systems.
The findings form part of Misplaced Trust: Inside the Workplace’s New AI Security Blind Spot, a new report from credential security leader Dashlane, which surveyed 3,000 knowledge workers and 800 IT security executives across the UK, US and France. It found a striking disparity between security executives’ perception of AI risk and the considerable autonomy that AI is being granted.
Nearly half (48%) of UK workers allow AI tools or agents to perform at least one activity without seeking permission first, including accessing emails, company files and customer information. One in ten (10%) allow AI to process payments without prior approval.
Employees are bypassing established security practices when using AI
More than a third (37%) of British workers would be willing to lend their login credentials to a new AI tool after just one week. Perhaps more concerningly, 45% of UK employees say they would comply if an AI tool asked them to confirm their login credentials every day, potentially creating opportunities for credential harvesting, impersonation and other forms of social engineering.
Only a quarter (25%) of UK employees say they always consider security risks when using an AI agent at work. Meanwhile, 15% of UK employees have granted an AI tool access to a file, system or account they subsequently discovered it should not have had. Among those who granted improper access, 43% believe that access remains active.
Nearly one in four employees surveyed (23%) have copied confidential, sensitive or proprietary information into
AI that they would not normally share with a colleague, website or other work system. A further 21% have skipped checking where information would be stored or sent, while 18% have shared personal or customer data in circumstances where they would not ordinarily do so.
Christophe Frenet, Chief Product Officer at Dashlane, commented: “The risks of AI spread faster, appear inside already approved tools, and unfold often without any further action from employees. This raises the stakes for organisations when employees grant access and expose sensitive data to AI tools or, as we’re seeing in our data, trust their new AI colleagues more than they would their human coworkers.”
Security leaders’ confidence is at odds with employee behaviour
The research also identifies a significant gap between the confidence of corporate security teams and the reality of AI adoption across British workplaces. While 92% of UK IT security executives express confidence that they know about every AI agent operating within their organisation, a quarter (25%) of employees admit to using AI tools their employer does not know about.
Some 44% of British workers have used AI tools for work in the past year without prior approval from their IT or security team, while 46% started using AI before their employer had established any rules governing its use.
Security incidents involving AI agent access are already being reported. Among UK IT security executives surveyed:
- 34% have experienced an AI agent accessing data, systems or credentials it was not authorised to access.
- 32% have experienced an agent taking an action it was not authorised to perform.
- 22% report unauthorised access to systems or data through an AI agent’s credentials.
- 9% say an AI agent’s credentials have been stolen, misused or compromised.
Overall, 72% of IT and security executives say their organisations have experienced a security incident involving an AI agent’s credentials or access in the past 12 months.
These behaviours raise questions about how effectively businesses can control and revoke permissions granted to AI agents.
Frenet commented: “The new findings indicate that AI-related access and identity risks are already affecting British organisations. Our latest data shows an emerging ‘AI Colleague Effect’ where employees are overextending trust to AI systems, sharing credentials and granting access to AI agents, sometimes without fully understanding what those tools can do with the permissions they’ve been given. That creates significant challenges for security teams trying to detect misuse or investigate incidents.”
Internal pressure to deploy AI quickly is leaving UK business exposed
The research suggests that the pace of AI adoption is placing further strain on security teams. More than two-thirds (68%) of UK IT security executives feel caught between pressure to deploy AI quickly and their responsibility to keep their organisation secure.
Among those facing pressure to adopt AI, 54% say this has resulted in faster deployment with less security review, while 35% have felt pushed to approve AI tools they were not confident in. Employee training and guidance are also failing to keep pace. Some 42% of UK workers say their employer has never provided guidance on what information is safe to share with AI, while only 37% have received security training focused specifically on AI-related risks.
Frenet concluded: “The challenge for security leaders is that the pace of change presented by AI and the potential consequences of misuse is unlike any other technology before it. The principles that make up the foundation of a good security program for securing human identities — just-in-time provisioning, least privilege, good credential hygiene and authentication — have to be applied to our new AI colleagues as well.”
