Protect agents using security capability
Security protects agents as and when new attacks surface. As agents gain access to critical data and systems, they become high-value targets. Hence security is critical to protect agents, identities, and data from misuse, oversharing and emerging AI threats, without slowing innovation.
Challenges
Without security, organizations face significant challenges, such as:
What access policies are enforced to block or restrict high-risk agent identities?
What guardrails are in place to prevent agents from accidentally oversharing sensitive data?
Is there a mechanism in place to detect and respond to new AI agent threats, such as tool misuse?
Solution
Security is at the heart of Agent 365’s capabilities. As agents become more capable and integrated into business processes, they also become attractive targets for attackers. Agent 365 delivers comprehensive protection for agents and data they use from external and internal threats. This allows you to innovate with agents confidently, knowing that risks are managed proactively and that your organization’s data and processes remain secure.
The following capabilities help organizations to secure agent identities:
Control agent access and start secure - Protect agent identities and prevent breaches by extending conditional access and identity protection from users to agents.
Apply granular access control to protect agent identities - Extend conditional access policies (including risk-based conditional access) from users to agents operating on their behalf; for example, a risky user giving a task to an agent that operates on their behalf can trigger multifactor authentication before an agent can proceed.
Extend conditional access policies (including risk-based conditional access) from users to agents operating on their behalf; for example, a risky user assigning a task to an agent that operates on their behalf can trigger multifactor authentication before an agent can proceed.
Limit agent access to only the resources they need - Assign access packages to agents after they're created to ensure they only have access to the necessary resources.
Prevent oversharing and data leaks - Gain visibility into AI-related data exposure, protecting the data agents create and access from oversharing, leaks, and risky behavior.
Defend against AI threats and vulnerabilities - Protect agents from threats and vulnerabilities, and adversarial attacks. Detect, investigate, and remediate incidents quickly, with visibility into attack paths.
