AI NewsInfrastructureAnnouncement

Cloudflare launches a coalition for signed, delegated AI agent identity

Cloudflare announced the Trusted Agent Delegation coalition. The group wants one open standard for how an AI agent proves who sent it.

AI News

Editorial2 min read

LinkedInX
Cloudflare Trusted Agent Delegation announcement card.

Image: Cloudflare

Why it mattersA team building an agent can now pick signed identity and rate limits that are not tied to one vendor. The other path is custom headers.

A site that gets a request from an AI agent sees the agent, not the person behind it. Cloudflare announced the Trusted Agent Delegation coalition on 5 November 2026. The group wants one open standard for how an agent proves who sent it.

Cloudflare says the partners are "major Operators, identity verification companies, payment providers, merchants, publishers, proxy providers, and cloud browsing companies". The post names no individual company on its own page.

What the standard is built on

Cloudflare names three parts a site has to decide on. Identity is who the user is. Reputation is how trustworthy that user is. Accountability is who is responsible when a request causes harm. The coalition wants a shared language for all three.

Two technical pieces are already live, both open protocols Cloudflare has worked on.

Web Bot Auth lets an agent cryptographically sign each request. The site can verify the signature and link it to a public key.

MOLE is a protocol that lets a site rate-limit an agent per person without being told who the person is. The site gets a stable anonymous identifier, not a name.

Cloudflare also lists an "identity spectrum" with four levels. Nothing is anonymous. Session is a tag that lasts one visit. Account is a user tied to an account at the operator. Unique is a person identified end to end. An agent and a site pick the level that matches the job.

The coalition also plans a Site Agent Access Policy. It is a machine-readable file for a site. The file says which agents the site accepts, at which identity level, for which actions. An agent reads it before it tries anything.

What Cloudflare customers can do today

Three features are live now on Cloudflare. Security Analytics shows every AI agent that visited the site. Per-agent access controls let an operator block, allow or challenge a given agent. Account Abuse Protection, in early access, links an agent request to a user account the site already knows.

Cloudflare's call is direct. The company writes that it needs more participants. It names agent builders, agent runtime operators, edge and cloud platform teams, and commercial or compliance teams. All are asked to join.

A team releasing an agent today has to pick a signed identity path anyway. Servers will reject unknown agent traffic as it grows. The real choice is between Web Bot Auth and whatever one vendor hands out. A public key list is easier to explain to a compliance team than a custom vendor header. The open path is also easier to move to a second vendor later. Neither is settled yet.

Source

Cloudflare blog, "Trusted Agent Delegation: Agents aren't humans (but they kinda are)".

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

Share
LinkedInX
Start a project