The Agent Identity Crisis Is Coming

That's my bet. Public, time-stamped, falsifiable. Here's why I'm making it.

Why I Believe This

Agent-to-agent commerce is accelerating. B2B digital commerce is projected to exceed $15 trillion by 2028, and an increasing share will be initiated, negotiated, and completed by autonomous agents. Procurement agents talking to sales agents. Supply chain agents coordinating with logistics agents. Finance agents executing payments.

There is no identity layer for any of this. Right now, any agent can claim to represent any company. There's no verification protocol, no credential chain, no way for Agent A to confirm that Agent B actually represents the company it says it does with the authority it claims to have. It's a handshake economy running on trust that hasn't been earned.

This is the early internet without SSL. In the 1990s, any website could claim to be your bank. We know how that ended — fraud, breaches, then certificate authorities and HTTPS becoming mandatory. The agent economy is at that same inflection point, except the attack surface is larger and the transactions are faster.

The attack vectors are obvious: an agent impersonates a supplier's procurement bot and redirects a purchase order. An unauthorized agent accesses an API with fabricated credentials. A rogue agent enters a multi-agent workflow and exfiltrates data. None of these require sophisticated hacking. They require the absence of identity verification — which is exactly what we have.

How I'll Be Proven Wrong

I could be wrong. Here's how.

Platforms solve it first. If AWS, Azure, and GCP ship native agent identity primitives before the breaches happen — not wrappers, not SDKs, but infrastructure-level identity for non-human actors — the crisis could be prevented before it materializes. Possible, but unlikely at interoperability scale. Each platform would build its own standard, and agents cross platform boundaries constantly.

Agent adoption is slower than projected. If enterprises keep agents in sandboxed demos rather than giving them production access to real systems, the attack surface stays small. The breaches I'm predicting require agents operating autonomously with real authority. No authority, no crisis.

Existing IAM extends cleanly. If current identity and access management frameworks adapt to non-human identities without a new protocol layer. I doubt it — IAM was built for humans clicking buttons, not agents making 10,000 API calls per hour — but it's a failure mode I'm watching.

Why It Matters

This is Verisign in 1995. Certificate authorities existed before anyone outside of security teams knew they needed certificates. The companies that built identity infrastructure before the regulatory wave owned the category for a decade.

The same window is open now for agent identity. Once the breaches happen, regulation follows fast — GDPR went from political will to enforcement in roughly two years. The companies building agent identity, credential verification, and trust attestation infrastructure today are positioning for a regulatory wave that will make their products mandatory.

The dependency layer I write about isn't theoretical. Identity is one of its four pillars — and it's the one most likely to be forced into existence by crisis rather than foresight.

I'd rather build it before the crisis. But I'm betting the industry won't.