Millions of AI agents are running unsupervised, survey warns

Gravitee says 2.4 million AI agents at major US and UK firms are operating without human oversight, based on a survey of 750 Chief Technology Officers and Vice Presidents of Engineering.

Respondents reported 7,239,650 AI agents deployed across businesses in the two markets, of which 2,445,553 were fully autonomous. Gravitee defines these as systems that can make decisions and take action without approval.

The research suggests use of these tools is rising rapidly. According to Gravitee, the number of AI agents appears to be doubling every six months. It also says there are now more AI agents working in the US economy than teachers, doctors, and lawyers combined.

A separate finding suggests governance has not kept pace with adoption. Just 7% of organisations surveyed said they had named a single accountable individual for the behaviour of AI agents.

Among those that had assigned responsibility, there was no standard for where that role sits. Some organisations pointed to the Chief Information Officer, while others named the Chief Data Officer or Chief Product Officer.

Pressure to deploy

The survey found that eight in ten organisations felt pressure to deploy AI agents even when security was not fully in place. For more than a third of respondents, that pressure came from the boardroom.

Technology leaders also expressed unease about the speed of adoption. Nearly a third said they were personally uncomfortable with the pace at which their organisation was moving.

The findings add to a wider debate over how companies are introducing autonomous software into core business processes. AI agents are designed to carry out tasks with limited intervention, but the degree of their autonomy and access to company systems has become a concern for executives, regulators, and security teams.

The survey also linked AI deployment to workforce plans. Nearly a third of firms, or 31%, said they were deploying AI agents specifically to reduce headcount.

That comes as employers in the UK and US face growing questions about how automation may affect jobs. Companies have increasingly framed AI software as a tool for efficiency, while workers and policymakers have raised concerns about substitution, accountability, and risk.

Gravitee commissioned the polling among senior technology executives in the US and UK. The respondents were Chief Technology Officers and Vice Presidents of Engineering, placing the findings among leaders directly involved in software and infrastructure decisions.

Control concerns

Rory Blundell, Chief Executive Officer of Gravitee, commented on the findings in stark terms. "There are now millions of AI agents loose at major firms: a number that's increasing every second. But what worries me is that a huge number of these are acting right now, without any oversight and with no accountability. There is nothing standing between them and causing untold chaos - leaking data, spending money, deleting files. C-suite executives need to get a grip on the agents that are being deployed in their business," he said.

His remarks reflect a line of argument becoming more common among software vendors and industry advisers: the challenge is no longer whether businesses will adopt AI agents, but whether they can monitor and control them once deployed.

For large organisations, that question cuts across cyber security, compliance, and operational risk. An autonomous agent with access to internal systems may be able to trigger transactions, alter records, move data, or interact with customers, raising questions about who is responsible when those actions go wrong.

The survey results suggest many companies have yet to settle that issue internally. With only a small minority naming one accountable person, responsibility appears to remain spread across technology and product teams.

Gravitee, which focuses on managing and securing digital infrastructure for businesses, released the research as scrutiny of AI governance intensifies. Boards and senior executives are under pressure to show they can adopt new tools without losing visibility over how those tools behave.

The headline figure is likely to draw attention because it suggests autonomous systems are already active at scale inside major organisations. Of the more than 7.2 million AI agents counted by respondents, about one in three were said to be operating with no human in the loop.