AI expansion creates security breaches: 75% of companies do not monitor AI agents

The rapid adoption of artificial intelligence by companies is creating a dangerous blind spot: supervision of the accounts of the AI agents themselves. A new Netwrix report reveals that 75% of organizations do not adequately monitor the activities of these accounts, and the consequence is a four-fold higher data breach rate among companies that use the most AI.

Why are AI agents accounts the critical blind spot in security?

Unlike human users, AI agents operate with non-human identities — service accounts, tokens and API keys that often have high permissions. The Netwrix report highlights that 75% of organizations do not have full visibility over these accounts, leaving them out of traditional security controls. This lack of supervision is aggravated by the fact that many IT teams still treat AI agents as software extensions, without applying the same identity governance policies used for employees.

How does lack of visibility in AI agents amplify the risk of rape?

The figures are alarming: the data breach rate in companies that use AI widely is 43%, compared to only 11% in those that do not use technology. This means that for every violation in a company without an AI, there are almost four violations in AI-intensive organizations. The attackers are exploring precisely the invisibility of these accounts to move laterally on the networks and access sensitive data. Recent cases of self-replicating AI worms in open models show how this attack vector is professionalizing.

What can Brazilian companies do against loopholes in AI agents?

First, it is essential to map all the non-human identities in the organization — from automation bots to generative AI assistants. Then apply the same principles of minimum privilege and continuous monitoring that are already standard for human accounts. Identity governance tools and detection of behavioral anomalies can help identify suspicious accesses before they turn into serious incidents. Netwrix also recommends periodic audits and implementation of multifactor authentication for AI agents whenever possible.

Frequently Asked Questions about AI agents and safety

What are AI agents accounts?

These are non-human identities used by artificial intelligence systems to access databases, APIs, and other resources — such as service accounts, tokens, and API keys, which often operate without direct human supervision.

Which sectors are most affected by this security gap?

Sectors with high adoption of AI, such as technology, finance, health and retail, are the most exposed, but any organization that implements AI agents without identity governance is at risk.

How do I start monitoring AI accounts in my company?

The first step is to carry out a complete inventory of all non-human identities, review their permissions and implement continuous activity monitoring tools, integrated into the existing SOC.

Source: [Utility Dive] (https://www.utilitydive.com/news/ai-identity-security-visibility-gaps-netwrix/822502/)