Watch this practical conversation about AI agents, what they actually are and why identity is quickly becoming the next big security question.
Right now, most teams are still trying to get their arms around the basics. In Latio’s latest Cloud Security Market Report, 65% of respondents said AI posture management is a top priority. That tells you one thing clearly: organizations want visibility into where AI is showing up, what models are in use and how these systems are connecting into real environments.
But visibility is only the first chapter.
They also report found 47% of respondents are also prioritizing access management. That matters because AI agents do more than sit in dashboards. They can call APIs, chain tools together, retrieve data and take action across systems. Once that starts happening, the real questions shift fast:
- What can this agent access?
- Which identity is it using?
- How long should access last?
- Who approved it?
- Can it escalate privileges?
There’s also a business reality underneath all of this. More than 65% of teams expect cloud security budgets to stay flat or decrease. So this is not about buying more tools. It’s about using smarter controls, reducing unnecessary standing access and bringing accountability to how agents operate.
In this recorded session, James Berthoty of Latio and Shashwat Sehgal of P0 Security start with a clear working definition of an AI agent, separate hype from reality and show where traditional IAM models start to strain. Then they get practical about authorization: how to decide what an agent should be allowed to do, in what context and for how long.
If your company is exploring AI or already putting agents into production, this will give you a useful framework and clear next steps.
Frequently asked questions
What are the core security risks of agentic AI in the enterprise?
The core security risks of agentic AI in the enterprise are over-permissioned agents and a lack of visibility into what actions an agent actually takes.
How do you govern and audit access for autonomous AI agents?
Teams govern and audit access for autonomous AI agents by applying least-privilege, just-in-time authorization and logging every action for review.
How does least-privilege, just-in-time access reduce agentic AI risk?
Least-privilege, just-in-time access reduces agentic AI risk by scoping every agent action to the specific task, removing standing permissions an attacker could exploit.
How does P0 Security secure AI agent access without slowing adoption?
P0 Security secures AI agent access without slowing adoption by automating authorization decisions instead of adding manual review steps.
