ID.me LLC

11/19/2026 | Press release | Distributed by Public on 11/19/2026 10:35

Why IDMe Joined the Open Secure AI Alliance

AI agents will soon act on our behalf. Trusted identity can help secure high-stakes workflows.

IDMe is now a member of the Open Secure AI Alliance, a Linux Foundation project that builds and shares an open defensive stack of AI models, tools, and techniques to secure software and AI agents. Personal AI agents can significantly improve the quality of life, from saving time and money executing tasks to discovering opportunities to enjoy the life you might have otherwise missed. At the same time, solving security concerns for people and organizations alike is a precondition to unlock this new future.

With more than 100 million Americans holding an IDMe Wallet and over 95 million verified to federal standards for trusted identity, IDMe brings experience in reusable, verified identity to the Alliance's work. This includes exploring how verified human identity, where appropriate, could complement secure execution environments such as NVIDIA's OpenShell. Identity and execution controls address different parts of the agent security challenge.

Agents are about to act for people

AI is moving from answering questions to taking action.

Picture Maria, an Army veteran who just took a new job and moved her family to another state. Today, that move means weeks of logins, forms, and phone calls: updating her address with the VA, transferring her kids' health records, filing taxes in two states, applying for a new driver's license, and setting up her new employer's onboarding.

Soon, Maria could hand much of that to an AI agent. She asks it to pull her W-2s and last year's return, request her kids' records from their old pediatrician on her behalf as their parent, check which state benefits she qualifies for, and fill out the paperwork for each one. The agent works across a dozen systems while Maria gets on with her week. It checks in only when something needs her approval.

That's the promise. But every one of those systems now has to decide whether to let the agent in.

That raises hard questions for every organization on the other end. Is this agent acting for a real, verified person? Did that person consent? Is the agent staying within the limits it was given? Can you audit it afterward?

Without trusted answers, organizations face two bad options: block agents and turn away the people using them, or let them through and open new paths for fraud.

There's a better path. Every agent should be able to prove three things:

  • Who it acts on behalf of: a real person, verified to federal standards.
  • That the person agreed: explicit consent, given by the person.
  • What it's allowed to do: a defined task, with limits and an end date.

When those can be checked at the moment of action and reviewed afterward, organizations can say yes to agents with confidence.

Security has to cover the whole agent stack

Much of the AI safety conversation has focused on the model. The Open Secure AI Alliance takes a broader view. An agent is a full system: models, the harnesses that let it act, guardrails, runtimes, enforcement points, and recovery. The Alliance's framework names "Policy, Identity & Governance" as one of the layers agent security depends on.

We agree. An agent that can't prove who it works for is a risk, no matter how well the model behaves. And that layer should be built in the open. Agencies, health systems, and enterprises need to inspect, test, and audit how agents are authenticated and authorized.

Bring agents onto trusted rails

Agents should use the trusted identity rails that already exist, not parallel, unverified pathways around them.

IDMe already holds a direct, verified relationship with the people we serve, and it's trusted where the stakes are highest. Today, 20+ federal agencies, 50+ state agencies, and 85+ healthcare organizations rely on IDMe for identity and sign-in. These are some of the most sensitive services an agent could touch, which is why trust and security have to come first.

The good news is that trust doesn't have to start from zero. Tens of millions of people could authorize an agent without proving who they are again. Organizations already connected to IDMe could accept agent authorizations, building on the integrations they already have today. And because fraud rings reuse identities, devices, and tactics across organizations, the signals IDMe connects across its network can help spot agent misuse that no single organization would see on its own.

We're building toward a model where:

  • The person starts it. A verified person delegates a specific task to an agent and approves it with the strong sign-in they already use. The authority to act comes from them.
  • The agent carries proof, not a profile. It receives a scoped token showing who it acts for, what it may do, and that the person consented. It shares only the facts the task needs. Share a fact, not your file.
  • The service knows it's an agent, and who sent it. The agency or business on the other end sees which verified person authorized the agent, through the IDMe integration it already uses, with a clear record of what was authorized.
  • The person stays in control. People should be able to see what they've authorized and take it back.

We're designing this on open standards like OpenID Connect and OAuth, including delegation models such as token exchange, so organizations can verify it with the tools they already trust and don't have to weaken their security controls to support agents.

Verify once. Carry that trust across participating services. Keep a clear record of who authorized what.

It has to work for everyone

Agentic AI can't become a tool that only works for people who are easy to verify. Our public sector partners serve everyone who is eligible, including people with thin credit files, unstable housing, or limited internet access. IDMe offers three ways to verify: online self-service, video chat, and in person at 700+ locations. Across all three, more than 9 in 10 legitimate users who attempt IAL2 identity verification pass. That same commitment to access will carry into how agents are authorized.

Why open, and why now

The rules for how agents prove who they work for are being written now. Standards set early tend to stick. Identity providers, the organizations that rely on them, and security teams need to shape those standards together, in the open, where the work can be inspected and improved.

That's why we joined. We'll bring what we've learned verifying identity at national scale to a community focused on securing AI for defenders everywhere.

Let's prove it on real workflows

The agentic future will be defined by the workflows we prove out first. We want those to be the services people depend on most: taxes, benefits, healthcare, and employment.

If you lead identity, security, or digital services at an agency, health system, or enterprise, we'd welcome a conversation. Reach out to your IDMe account team to scope a pilot where agents act on trusted, auditable identity rails.

Strong, inclusive, privacy-protecting identity has always been our mission. The agentic era doesn't change that. It raises the stakes.

Blake Hall is the Co-Founder and CEO of IDMe.

ID.me LLC published this content on November 19, 2026, and is solely responsible for the information contained herein. Distributed via Public Technologies (PUBT), unedited and unaltered, on November 19, 2026 at 16:35 UTC. If you believe the information included in the content is inaccurate or outdated and requires editing or removal, please contact us at [email protected]