At Identiverse 2025, identity experts explored the future of AI agent identity and access management. Key challenges include authenticating agents, governing their lifecycle, and securing autonomous interactions. Microsoft Entra is pioneering solutions to ensure safe, transparent AI agent governance.
AI Agents and the Future of Identity: Insights from Identiverse 2025
At Identiverse 2025, identity pros worldwide gathered to tackle a hot topic: how do we manage AI agents in identity and access management (IAM)? Microsoft Security hosted a roundtable with 149 experts from finance, healthcare, government, and tech. The goal wasn’t to sell solutions but to explore challenges and shape the future of securing AI agents.
What’s New: AI Agents in Identity Management
AI agents are no longer sci-fi—they’re embedded in daily workflows. From summarizing meetings to automating helpdesk tasks, these agents are getting smarter and more autonomous. This raises big questions for IAM:
- How do we authenticate and authorize AI agents securely?
- How do we govern their lifecycle and prevent shadow identities?
- How do we distinguish agents from humans?
Microsoft is actively collaborating with customers and partners to develop new solutions for these challenges.
Major Updates from the Roundtable Discussions
Defining AI Agent Identity
Experts debated what “identity” means for AI agents. Are they service principals, workload identities, or something new? Consensus: agents need persistent, first-class identities. Some treat agents like users with licenses and delegated permissions, while others mix user context with agent attributes. As one participant put it:
“We’re seeing agents expose the lack of control that already exists. They’re not creating new problems—they’re surfacing old ones.”
Authentication and Authorization Challenges
Securely authenticating agents and assigning roles proved complex. Participants want agent-specific access controls, scoped permissions, and behavioral monitoring. Current IAM systems aren’t fully ready, signaling a need to rethink standards, especially for multi-factor authentication (MFA).
“We need to rethink how MFA works when an agent acts on behalf of a user.”
3. Security and Governance Concerns
Agents can amplify existing vulnerabilities like over-permissioned data. Experts stressed containment strategies such as dynamic permissioning and anomaly detection. Governance gaps remain, especially around agent ownership and lifecycle management. The consensus: manage agents like humans—onboarding, offboarding, and all in between.
4. Agent Experiences and Transparency
How do admins discover agents? How do agents find each other? Transparency is key, with calls for directories, visual indicators, and audit logs. Preventing impersonation and designing user-friendly agent interactions are top priorities.
“We don’t just need to govern agents—we need to design for how they’re experienced.”
What You Need to Know
The identity community is still defining IAM for AI agents. Shared language, frameworks, and responsibility are essential. Microsoft Entra is leading with a unified directory for agent identities and plans to release more capabilities soon. If you’re tackling these challenges, join Microsoft Entra Agents cohorts to learn and contribute.
Stay ahead in AI agent security by following Microsoft Entra updates and joining the conversation in the Tech Community.
From the Microsoft Entra Blog articles
