AI Agents Are Joining the Workforce: Why Identity Governance Needs to Keep Up
As AI agents become more common in the workplace, securing them is becoming a serious identity challenge. At SailPoint Navigate in Sydney, Chris Cubbage of Australian Cyber Security Magazine spoke with SailPoint Chief Product Officer Levent Besik about the issue. Besik described three key challenges: the sheer scale of non-human identities, the complexity of agents acting with human credentials and access, and the speed at which risks can unfold.
The discussion covered why sandboxing alone is not enough to keep agents safe. Besik stressed the importance of assigning clear ownership to every agent, so someone is accountable for what it can access and do. He also outlined a three-stage approach to agent governance: discover, govern, and protect. In practice, that means knowing which agents exist, controlling what they can reach, and defending against misuse.
The conversation also touched on real-time authorization, preventing sensitive data uploads, and integrating with existing security tools. Another topic was setting guardrails around high-impact or irreversible actions. This summary is based only on the opening of the article, so further detail from the full interview is not covered here.