Cloudflare says sites need identity tools for users behind AI agents
Original title: Agents aren’t humans (but they kinda are)
Remotely hosted agents hide the user, so sites lose the identity, reputation, and accountability signals they use for fraud and access.
Cloudflare argues that operator-hosted personal AI agents act as intermediaries, so a site sees the agent rather than the user behind it. It says blocking intermediaries worked historically but fails if most traffic becomes agentic. The post says identity, reputation, and accountability must still bind users, agents, and operators. Agents running on the user's own device, including OpenClaw, are outside the main scope.
Unconfirmed: The claim that agentic traffic may eclipse human browsing is the author's projection, not a measured figure.
- Agents are automations that pursue a user's goal on a cloud computer or browser.
- Examples named are Muse from Meta and dots from OpenAI.
- Named problems include account sharing, age or geographic checks, and friendly fraud.
- The Intermediation Principle says user-site duties must survive intermediation.
- Local and open-source agents on the user's device are out of scope.
- Organizations
- CloudflareMetaOpenAI
From the original article
14 min read at blog.cloudflare.com
AI agents act for millions, but servers see intermediaries that strip user visibility. This post defines the problem space and introduces a collaborative approach to safely identifying, allowing, and managing this automated traffic.
By Bryan Becker · blog.cloudflare.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.