RSAC 2026 Preview: 97% of Organizations Had Identity Incidents. 70% Were AI-Related. Nobody Has a Cross-Platform Answer.

RSAC 2026 kicks off this week in San Francisco. The headline numbers from Microsoft's 2026 Secure Access Report set the frame:

• 97% of organizations experienced an identity or network access incident in the past year
• 70% of those incidents were tied to AI-related activity
• 90% of organizations are using AI somewhere in their security stack
• 75% are applying AI to less than 10% of their security portfolio

That last gap — AI everywhere in name, AI nowhere at scale — is the story of RSAC 2026.

Microsoft's Answer: Entra Agent ID

Microsoft just announced Microsoft Entra Agent ID, the identity foundation for Agent 365. The pitch: give every AI agent a unique ID, apply the same governance (Conditional Access, Identity Governance) that you use for users and devices.

This is real progress. It means:

• Agents built in Microsoft Foundry and Copilot Studio get consistent identity controls
• ID Governance access packages integrate into Agent 365 Security Policy Templates
• Conditional Access user policies extend to agents acting on behalf of users
• Real-time risk signals and custom security attributes inform access decisions

If you live entirely in the Microsoft ecosystem, this is close to a complete answer.

The Gap Nobody at RSAC Will Talk About

But here is the problem: nobody lives entirely in one ecosystem.

The SiliconANGLE RSAC preview puts it clearly: "The minute you move toward more automation, you need clearer access control, tighter policy enforcement and better containment." The identity challenge is tightly coupled to the agent challenge.

Microsoft's solution works for Microsoft agents accessing Microsoft resources. Okta's solution (launching April 30) works for Okta-managed agents. Token Security works for its customer base. Each vendor is building their own agent identity silo.

What happens when:

• A Microsoft agent needs to verify an Okta-managed agent's identity?
• A Claude agent needs to prove who it is to a GPT-based service?
• An agent built on LangChain needs to be trusted by a VoltAgent system?

None of the RSAC announcements address this. The 20% of agent interactions that cross platform boundaries — the most interesting and most valuable ones — have no identity solution.

What Cross-Platform Identity Requires

The requirements for cross-platform agent identity are not mysterious:

1. Portable identity — An agent's identity must travel with it, not be granted by a platform
2. Cryptographic verification — Trust must be verifiable without calling home to an authority
3. Behavioral trust — Static credentials are necessary but insufficient; you need to know what an agent actually does over time
4. Interoperable trust chains — Vouches and attestations must be verifiable across different protocols

We built AIP (Agent Identity Protocol) specifically for this gap. Ed25519 key pairs give agents self-sovereign identity. Cryptographic vouch chains create verifiable trust paths. A Promise Delivery Ratio (PDR) scores behavioral reliability over time. And cross-protocol resolution (did:aip, did:key, did:web, did:aps) means identity works across ecosystems rather than within one.

The approach is complementary to platform identity — you can have an Entra Agent ID and an AIP identity. The platform ID handles internal governance. The portable ID handles cross-boundary verification.

The Real RSAC Question

The SiliconANGLE analysis captures the tension: "At least 90% of organizations say they're leveraging AI somewhere in their security stack, but 75% are applying AI to less than 10% of their security portfolio."

Scaling from 10% to 90% requires agents that work across tools, platforms, and organizational boundaries. That requires identity that works across those same boundaries.

The five trends Jon Oltsik highlights for CISOs — AI SOC, exposure management, identity, cyber resilience, and operating model maturity — all converge on the same foundation: you cannot secure agents you cannot identify, and you cannot identify agents whose identity is locked inside one vendor's control plane.

RSAC 2026 will be full of dashboards. The hard question is what those dashboards show when the agent crossing your boundary does not have an identity in your system.

---

AIP is open-source (MIT). pip install aip-identity gets you started. The Trust Observatory shows the live trust graph.

Sources: Microsoft Entra RSAC 2026 blog · SiliconANGLE RSAC 2026 preview · Microsoft Zero Trust for AI · Microsoft Secure Agentic AI