CyberArk Introduces the first solution to Protect AI Agents

The adoption of artificial intelligence in corporate processes is happening at incredible speed. According to Gartner forecasts, by 2028, at least 15% of day-to-day work decisions will be made autonomously using AI agents (compared to 0% in 2024). However, this opens up a new attack vector.

To perform tasks, AI agents require access to sensitive data, applications, and cloud resources. Essentially, they gain the same rights as human users but operate at machine speed and scale. CyberArk has introduced a solution to this problem.

Security purpose-built for AI

Traditional access controls are often not flexible enough for the dynamic nature of artificial intelligence. The new CyberArk Protect for AI Agents solution extends the capabilities of the Identity Security platform, providing organizations with tools for the secure adoption of AI technologies.

This solution allows applying the principle of least privilege to AI agents, preventing the possibility of their identities being used for cyberattacks.

Key solution capabilities

Full visibility

• Automatic discovery of AI agent identities.
• Classification of access levels and rights.
• Real-time activity monitoring.

Zero standing privileges

• Implementation of the Zero Standing Privileges approach for machine entities.
• Provision of "Just-in-Time" access.
• Automatic revocation of rights after task completion.

Anomaly detection

• Analysis of AI agent behavior.
• Blocking suspicious actions that fall outside standard patterns.
• Prevention of data leakage through compromised agents.

Human vs. AI: why a different approach is needed

Security models designed for humans are not always effective for machines. Let's look at the key differences in access management.

Human access

• Speed: Limited by physical input capabilities.
• Context: Often predictable working hours and location.
• Authentication: Multi-factor (MFA), biometrics.

AI agent access

• Speed: Instant processing of thousands of requests.
• Context: 24/7 operation, dynamic instance creation.
• Authentication: Requires automated rotation of secrets and tokens without human intervention.

Softprom is an official distributor of CyberArk. Our team will help you assess the risks associated with AI adoption and demonstrate the new solution in practice.