Agentic AI security pushes vendors toward shared safeguards

Agentic AI security is increasingly becoming a shared concern among industries as autonomous systems become more intertwined in various products and platforms. Okta Inc., a leading vendor in this space, emphasizes the need for coordinated safeguards rather than isolated vendor controls to build trust. Charlotte Wylie, Okta's senior vice president and deputy chief security officer, highlighted the importance of communicating their approach to a broader audience, beyond just security organizations and technology conferences.

Wylie stated that the industry should ensure that their messaging reaches families and friends of the attendees, sparking awareness about the measures being taken to secure agentic systems.

Okta is part of the Blueprint Alliance, a group comprising Amazon Web Services Inc., CrowdStrike Holdings Inc., Salesforce Inc., and ServiceNow Inc. The alliance aims to make security controls and risk signals interoperable across agentic systems. According to Wylie, the industry must work together to solve this problem and provide an architecture that can secure the entire agentic stack.

This architecture should assign each agent an identity, limit its permissions, and provide visibility into its behavior, enabling organizations to promptly detect any deviations from the agent's intended purpose. By implementing these controls, organizations can effectively govern permissions and stop rogue agents from causing harm, ensuring that technology can be used securely and efficiently.