Nearly three-quarters of organizations experienced a security incident involving an artificial intelligence agent's credentials or access during the past year, according to new research commissioned by credential security company Dashlane.
The report, "Misplaced Trust: Inside the Workplace's New AI Security Blind Spot," also found that employees are sharing sensitive credentials and granting AI tools access to corporate information, while IT and security leaders report challenges maintaining oversight of AI deployments.
The research surveyed 3,000 full-time knowledge workers and 800 IT security executives across the United States, United Kingdom and France.
Among the findings, 56% of workers acknowledged granting AI privileges they would not normally extend to a coworker or work system. That included 21% who shared passwords, API keys or login credentials, 21% who shared personal or customer data and 25% who shared confidential corporate information.
The survey also found that 55% of workers allow AI tools to perform actions autonomously, including processing payments and accessing emails and customer data.
AI adoption pressures security teams
The report identified pressure to accelerate AI adoption as a significant challenge for IT and security teams.
Sixty-one percent of IT and security leaders cited pressure to move quickly with AI as one of the biggest barriers to securing AI agents. Among respondents facing pressure to implement AI, 57% reported that it resulted in faster deployments with less security review.
Additionally, 53% of IT and security leaders said some AI tools are deployed without undergoing a security review, while 71% of organizations reported granting AI agents standing access to internal systems or data.
"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," said Christophe Frenet, chief product officer at Dashlane. "Balancing speed and security has always been an issue front and center for enterprises, but with AI, the stakes of getting it wrong are much higher."
Security leaders struggle with AI visibility
The research also identified differences between security leaders' confidence in their oversight of AI tools and their organizations' experiences discovering those tools after deployment.
Although 89% of IT and security leaders expressed at least some confidence that they knew about every AI agent in use, 98% of organizations reported discovering AI agents after deployment.
Meanwhile, 28% of workers acknowledged using AI tools without their employers' knowledge.
The findings also examined how employees respond to requests from AI tools compared with requests from human coworkers.
According to the survey, 48% of workers said they would share their login credentials with a new AI tool after one week, compared with 42% who would do the same with a new human coworker.
Another 52% said they would comply if an AI tool asked them to confirm their login credentials every day.
Accountability and oversight budgets
The report also examined responsibility for security incidents involving AI agents and expectations for additional security resources.
When asked about accountability for AI-related incidents, 69% of IT and security leaders said responsibility would fall on them, while only 11% assigned responsibility to the business unit that deployed the technology.
Organizations also anticipate additional demands on their security teams as AI adoption continues.
Eighty-four percent of IT and security leaders expect their budgets for AI agent oversight to increase over the next 12 months, while 79% anticipate an increase in their AI oversight workload.
Commissioned by Dashlane, the survey was conducted by Method Research and distributed by PureSpectrum from Aug. 18 to Sept. 5.


