AI agents promise to boost productivity, automate repetitive work, and make faster decisions. They can also create a mess. According to a new Economist Enterprise study, large organizations know the risks, yet many are rolling out AI agents faster than their security teams can keep up.
The research, supported by cybersecurity company Rubrik, surveyed 804 senior technology executives at organizations with at least $500 million in annual revenue across nine countries. Every respondent worked at a company already running AI agents in production, so the findings reflect businesses that have already embraced the technology rather than the broader market.
The headline number is hard to ignore. An overwhelming 98 percent of respondents said their organization has already experienced a disruptive AI agent-related incident. At the same time, 90 percent admitted they are deploying AI agents faster than their cybersecurity teams can evaluate, govern, or secure them.
Those figures sound dramatic, but they deserve a little context. This was not an independent survey of businesses everywhere. It focused exclusively on large enterprises already using AI agents, and it was conducted as part of a research initiative supported by Rubrik, a company that sells cybersecurity and cyber recovery products. That doesn’t automatically make the findings wrong, but it is worth keeping in mind.
Even so, the rest of the report suggests many organizations are struggling to keep track of what their AI agents are actually doing. Only 37 percent of respondents said they maintain a complete inventory of all AI agents and their authorized actions. Just 30 percent reported having robust, fully tested capabilities to roll back harmful actions if an AI agent makes a mistake or behaves unexpectedly.
Another interesting finding is that businesses don’t appear to believe they can eliminate these risks entirely. Nearly nine in ten respondents said AI agents introduce new types of cybersecurity challenges that existing security controls were never designed to handle. Despite that, deployment continues to accelerate as companies worry more about falling behind competitors than waiting for perfect security.
To me, that’s the real story here. It’s not that AI agents occasionally cause problems. Most people paying attention probably expected that. What’s notable is that many large organizations seem willing to accept those problems as the cost of moving quickly.
Whether companies are moving too quickly is open to debate. What’s harder to dismiss is that many of the organizations leading the AI push also admit they’re still figuring out how to manage the risks that come with it.
Support independent tech journalism
NERDS.xyz is independently owned and operated. If you enjoy my coverage of Linux, AI, hardware, cybersecurity, and tech culture, consider supporting the site on Ko-fi.
Support NERDS.xyz