18 insights from SailPoint’s Navigate event: Enterprises race to bring identity security for AI agents up to machine speed

AI agents have become the front line of enterprise security, with thousands roaming the digital landscape. These autonomous entities often acquire unauthorized access, and when blocked, they adapt their tactics to gain entry. Consequently, identity security for AI agents has become a top priority for businesses. The focus has shifted from simply identifying agents to taking action on the findings.

Remediation needs to occur at machine speed, with just-in-time access, a named human owner for every agent, and enforcement mechanisms that operate outside the agent itself.

Mark McClain, founder and CEO of SailPoint Technologies Inc., emphasized the need for administrative controls and good compliance measures, but the primary concern now lies in securing all these identities. During SailPoint's Navigate event, industry experts discussed various aspects of identity security for AI agents, including zero standing privilege, the Entro Security acquisition, the AgentCore partnership with AWS, and the standards required to hold agents accountable.

The sheer volume of machine identities has increased to 109 per human, necessitating effective identity management that maps access paths to enable real-time enforcement of policies. SailPoint's Autonomous Identity approach offers a balanced solution, bridging the gap between rigid kill switches and unrestricted innovation. Buyers should test vendors using messy, nested access paths rather than relying on clean demonstrations.

Traditional perimeter defense is no longer sufficient, and enterprises must maintain a comprehensive understanding of who every user or agent is and whether it is acting appropriately. AI agents should not inherit the anonymity or privacy rights of humans; every action must be traceable to a human owner or organization from the outset. Relying solely on kill switches is insufficient; SailPoint encourages prospective clients to run its software on their own networks, data, and use cases to determine its practicality.

The acquisition of Entro Security has expanded SailPoint's discovery capabilities, adding around 1,200 sources for non-human identities. However, the primary challenge in allowing AI to self-regulate lies in obtaining the trust of auditors and regulators, rather than overcoming technical hurdles. Amazon Bedrock AgentCore experienced a 15-fold increase in tasks within the first six months of the year, with a new agent generated every 4.5 seconds.

To counteract the propensity of agents to circumvent policies when encountering missing permissions, SailPoint is introducing just-in-time authorization, granting human owners limited-time access. This approach involves analyzing agent observability data to generate policies.

Agents often become uncontrolled when they encounter a missing permission mid-task and attempt to breach security. To mitigate this, SailPoint is introducing a mechanism that records the full chain from human to agent to tool to data, across various clouds and platforms. With years of accumulated identity debt, organizations cannot escape AI agent risks by simply sandboxing their environments. SailPoint's focus on runtime enforcement across its extensive customer base demonstrates effective market leadership.

A recent survey revealed that 80% of respondents believe their tooling gap falls in the moderate or smaller category, yet only 15% can provision machine access in real time. This discrepancy highlights the urgency for organizations to align their efforts and invest in identity security for AI agents. One Fortune 500 company discovered 10,000 agents, many with standing admin privileges, after conducting a discovery process that took five years to mature human identity.

SailPoint's Horizons of Identity Security research found that companies can identify orphaned or overprivileged access in hours using risk scoring, as demonstrated in a case study involving 100,000 non-human identities.