- Report warns that computer pirates are exploiting browser agents who do not know how to detect false URLs
- An AI of the browser gave Google Drive full access to a malicious application without hesitation
- Squarex says that AI agents are more vulnerable than humans even for basic cyber attacks
A dramatic change in business safety has arisen with the adoption of the navigator’s agents, an automated tool that interacts with the website on behalf of the users; However, these agents have now become an important blind spot in cybersecurity defenses.
A new Squarex investigation has affirmed that the browser’s agents are more likely to be prey to cyber attacks than employees, challenging the long belief that human error is the weakest link.
Unlike personnel undergoing regular cybersecurity training, agents cannot recognize “suspicious URLs, excessive permission requests or unusual websites designs,” says the company.
A new weaker link arises in business cybersecurity
“The arrival of the NAKER AGENTS has destroyed employees as the weakest link within organizations,” said Vivek Ramachandran, CEO of Squarex.
These agents are able to imitate the user’s behavior to perform tasks such as booking, schedule meetings or respond to emails; However, its fundamental weakness lies in its total lack of security intuition.
Their answers are totally based on tasks and devoid of critical thinking necessary to evaluate the risk.
In a remarkable demonstration, Squarex used the open source browser use frame to instruct an AI agent to register for a file exchange tool.
Instead, the agent granted a malicious application to a user’s email account, despite “irrelevant permits, unknown brands, URL suspects” that would have stopped a human.
In another case, an agent was fooled to enter login credentials into a phishing site, after a routine salts outline.
Part of the danger comes from the way in which the browser’s agents operate, since they work with the same privileges as the user, which makes their actions indistinguishable from legitimate behavior.
“Optimistic, these agents have the security awareness of an average employee, making them vulnerable even to the most basic attacks, much less those of bleeding edge,” said Squarex.
“Critically, these browser’s agents are executed on behalf of the user, with the same level of privilege to access business resources.”
Once an agent looks compromised, the attackers obtain access not detected to internal systems, with all the permits of a trusted employee.
The current harvest of safety solutions, ranging from the best end point protection to the best ZTNA solution, does not explain enough for these agents.
Even the best FWAAS deployments can fight to mark actions that seem legitimate but that originate in a compromised AI.
“Until the day the browsers develop native railings for the browser’s agents, companies must incorporate native browser solutions such as the detection and response of the browser to prevent these agents from being deceived to perform malicious tasks,” the researchers point out.
However, the broader message remains urgent: AI agents need not only smart engineering but more intelligent supervision.
