Research from identity management supplier Okta is pointing to new security problems, connected with the emergence and growth of artificial intelligence (AI) agents and non-human identities (NHIs).
Commissioned by Okta and conducted by research firm AlphaSights, the supplier’s AI at work survey polled 260 C-level executives across nine countries – Australia, Canada, France, Germany, India, Japan, the Netherlands, the UK and the US – in May 2025.
Okta said the research “uncovers a complex landscape where efficiency and speed are top priorities but security remains critical to realising AI’s full value”.
Stephen McDermid, EMEA chief security officer at the firm, said: “The rise of AI agents and non-human identities underscores the urgent need for new security and governance frameworks.
“AI adoption is more than a technical upgrade – to maximise its strategic value, organisations must define clear use cases and build solid foundations from the start,” he added.
“Whether it’s using AI for customer support interactions, sales and marketing initiatives or security enhancements, identity is fundamental to how businesses will optimise their use of AI, whilst also keeping their data safe.”
Governance and oversight
According to the research, the main security concern related to AI agents over the next three years is governance and oversight, cited by 58% of organisations, followed by compliance and regulatory requirements at 50%.
Securing NHIs is now a “critical” or “very significant” priority for 76% of organisations globally, with 78% citing control of NHI access and permissions as their leading concern.
However, only 10% of organisations have a governance strategy in place for managing NHIs, while 91% are already deploying AI agents.
In an interview with Computer Weekly, McDermid said the projected proliferation of agents in enterprise software is “terrifying”.
“You could have agents yourself, doing your shopping or trying to make your life easier in other ways,” he said. “But the challenge is when you have an agent talking to another agent. How do you define the boundaries between what they are and what they are not allowed to do?
“And we know from speaking to customers that they have the challenge now not only of building their own agents, but having them work with other agents, too,” said McDermid. “That is where it gets really messy and complicated, and that is why if you look at the exponential growth [of agents], people are responding with talking about governance and frameworks to avoid data being shared inappropriately, which is already happening in the field.”
The survey found that globally, AI adoption has surged from 64% to 80% in the past year, with 62% of organisations citing security and threat detection as the top AI goal for the year ahead.
In EMEA, AI adoption closely matches the global rate, but country-level variation is significant, according to the survey. France is leading with 83%, followed by Germany (77%), the Netherlands (70%) and the UK (70%). However, only 9% of global respondents described themselves as “very prepared” to defend against AI-driven threats.
Some 85% of global respondents now view identity and access management (IAM) as important to their security posture, up from 79% last year.
The share of companies using AI for security purposes decreased from 71% in 2024 to 65% in 2025, but the cohort of respondents who “strongly agree” with the view that “AI is the best defence against AI” rose from 18% to 41%.
In EMEA, the use of AI for security seems strong, with 70% of UK organisations exploring AI-driven security systems, and France (63%), Germany (57%) and the Netherlands (57%) also showing uptake. These figures still trail global leaders such as the US and India, said Okta.
Fast growth in AI agents and NHIs
In 2025, 42% of organisations globally attested to widespread use of NHIs. In EMEA, adoption is lower overall at 22.5%, with the Netherlands leading at 30%, and the UK, France and Germany each at 20%.
AI agents seem to be complementing human workforces, according to this research. The most common reported benefit from integrating with AI agents is increased productivity, followed by cost savings.
For 2025, 47% of companies expect AI agents to have a moderate impact on operational efficiency, and 55% believe these agents will complement human roles, while only 5% see AI agents replacing human workers. Top trends for AI agents include – unsurprisingly – automating repetitive tasks; enhancing customer service; and optimising workflows.