"The agents on such platforms operate autonomously and create new risks for data breaches. For example, employees can pass on sensitive information via uncontrolled prompts or install risky plugins. The new security layer gives security teams insight into which agents are active and blocks unauthorized data access. Since copilots and agents need to have access to sensitive data in order to be useful, such insight is necessary to ensure that access management is carried out as desired."
"The browser, for example, is anything but a secure domain, where AI-driven attackers can deceive unsuspecting employees just like with phishing emails. Think of contact forms that turn out to be fake in order to extract sensitive data, or hijacking a browser session to obtain cookies. Because cyberattacks are occurring more frequently via the browser than before, organizations need to find a way to protect this attack surface. Palo Alto Networks notes that 95 percent of organizations report incidents originating from the browser."
Prisma SASE 4.0 introduces SaaS Agent Security to monitor autonomous AI agents with access to company data, currently supporting Microsoft Copilot Studio and ServiceNow. The module reveals active agents, blocks unauthorized data access, and helps enforce access management because copilots and agents require sensitive data to function. Autonomous agents present risks similar to unpredictable employees, including uncontrolled prompts and risky plugins. Attackers increasingly use AI to target browsers through fake contact forms, session hijacking, and cookie theft. Organizations report browser-origin incidents frequently, and attackers bypass network controls, exploit interactive sessions, and weaponize DNS, prompting DNS-layer protections like Palo Alto's Advanced DNS Resolver.
Read at Techzine Global
Unable to calculate read time
Collection
[
|
...
]