Innovators Spotlight: Badge Unveiled – Part II

Badge Is Creating the Trust Framework Your AI Agents Have Been Missing

This piece builds on my earlier conversation with the Badge team at RSAC 2025, which led to a follow-up discussion at RSAC 2026 in San Francisco. You can find that original article here. Here, we explore how Badge is tackling one of enterprise security’s most persistent challenges — building portable, cryptographic trust that spans humans, machines, and AI agents — all without relying on a single stored shared secret.

Spend enough time walking the RSA Conference exhibition floor, and the same themes start echoing everywhere. Every vendor is pitching zero trust, AI integration, and “reinvented identity.” But very few can break down, in real cryptographic terms, how any of it actually functions once it lands inside a complicated enterprise — one shaped by years of mergers, partially migrated directories, and a growing pile of orphaned credentials.

Then you have a conversation with Dan Kaufman and the Badge team, and suddenly everything clicks.

Radiant Logic experienced this firsthand. As Dan explains, they approached Badge with what felt like an awkward admission: their customers were knocking on the door demanding something Radiant Logic simply didn’t offer.

“Radiant Logic came to us and said, word for word, ‘Hey, we’re getting all these [customers] with all this incoming data, and we have all these logos… all these logos are coming in asking exactly for that, and we see that you’re working with other people. Why not us?’ We said, ‘We weren’t deliberately avoiding working with you. We didn’t realize that’s what you wanted or what the opportunity looked like.'”

Once they aligned on the vision, the next steps were clear: embed Badge’s capabilities directly into Radiant Logic’s identity fabric so customers can access the functionality as a built-in feature rather than a separate add-on.

“We collaborated with them for a while, built the integration, got everything in place so they can simply go to market with it. And now we’re announcing that we have a partnership.”

For CISOs, that combination is compelling because of what each party contributes. Radiant Logic already operates at the heart of intricate identity infrastructures inside some of the world’s largest enterprises. Badge integrates not as yet another isolated identity system, but as what Dan calls a cryptographic backbone.

Dr. Tina Srivastava, co-founder of Badge Inc., summed it up neatly:

“It’s starting to position Badge, at least in the market, as this kind of trusted backplane — where the key question becomes, ‘Is that backed by Badge?'”

Peel away the marketing language, and that’s precisely the kind of question CISOs want their security architects asking: what is the root of trust behind this identity decision, and can I apply it universally — for humans, machines, and now agentic AI?

Agentic AI Deserves a Badge as Well

Badge began with human identity because that’s the most intuitive entry point for the industry to grasp biometrics and risk-based authentication. But the team is emphatic that the same cryptographic foundation extends seamlessly to AI agents.

“You wrote [last year] very clearly about how Badge lets you generate a key on the fly from any set of factors — whether that’s a human biometric, device signals, hardware, or machines. And so what we’re able to do is extend that coverage to your agentic AI as well, right? So why is it called agentic AI? It’s because you’re granting it the authority to act on your behalf. And so if you have a cryptographic key, you can now issue delegated, scoped credentials for your agentic AI agents to operate on your behalf.”

Consider what most vendors are actually delivering under the “AI assistant” label today. In the best case, you get a clever macro connected to an API token that can handle a limited set of tasks. In the worst case, you get an overprivileged chatbot with access it never should have been granted in the first place.

Badge is pushing the industry toward a more mature conversation. If an agent is going to act on a human’s behalf, it needs the same things a human requires in any serious security program:

• A cryptographically robust identifier anchored to a root of trust
• Delegated, scoped credentials that clearly define what it can and cannot do
• A mechanism for relying parties to confirm the agent was genuinely authorized for this specific action

Dan illustrates with a healthcare scenario that will make any privacy-conscious CISO cringe — largely because it’s so believable. Imagine you’re frustrated with your current insurer and tell your agent to transfer your records and coverage to a different provider.

“It’s a perfect task for agentic AI to handle, because you’d say, ‘Oh, please just take care of that. Switch me from one insurance company to another.’ Okay, now what happens? Here’s the problem…”

He then outlines three critical gaps that will resonate with anyone who has watched an AI proof of concept quietly morph into a risk register item:

1. “How does the relying party know this agent is genuinely acting for you, rather than just sitting on a pile of your data?”
2. “How do relying parties distinguish between multiple agents linked to the same person, each designed for different purposes — like medical versus shopping versus travel?”
3. “Even if identity and intent are established, how do you guarantee the agent stays within a strict scope and isn’t silently leaking data to third parties?”

“If you don’t have this trust layer embedded underneath, it’s going to be difficult. Right now it’s the wild west. Everyone’s racing ahead, it’s free-for-all, but once we dig in, you can think of a million examples.”

The bottom line is straightforward:

“Everything that applied to human identity — this exact same technology, already built, already proven — applies in this [agentic] world as well. We always say, ‘How do you prove you’re you?’ With an agent, it’s precisely the same question. How do I prove that the agent is legitimate?”

For an industry that loves chasing flashy models and “AI-native” platforms, this is a refreshingly pragmatic stance. Badge isn’t trying to compete as the best foundational model or the most eye-catching AI co-pilot. They’re unapologetically focused on ensuring you can trust whatever agent you ultimately deploy — no matter whose model powers it.

No Stored Secrets, Real Cryptography, and the M&A Headache

If you want to gauge how serious a vendor is about identity, ask them what happens after a merger. Most will start gesturing vaguely at roadmaps and “phased consolidation plans.” Badge talks about X.509 certificates, Kerberos, and the reality that your directory topology resembles a family tree sketched out by a committee.

Dan continues, “Healthcare is a great example, because healthcare has really expanded through consolidation. You frequently have disconnected networks. You have systems where gaining access is genuinely difficult — clinics merge, hospitals merge. So you end up with these fragmented systems.”

In that environment, Radiant Logic’s identity fabric handles policy and routing. Badge provides a way to authenticate across these patchwork domains without flooding the environment with passwords or spending the next three years synchronizing directories.

“Specifically, an example comes from Cerner Cloud, which requires access through an AD trust that supports only one AD domain. And so that becomes a challenge, especially in healthcare. We have all kinds of domains… Our joint solution now provides SSO from any managed AD domain into that Cerner AD domain and the authorization that goes with it.”