Agents aren’t humans (but they kinda are)
Web page · https://blog.cloudflare.com/trusted-agent-delegation
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.
First seen by Oriel on .
AI summary
Cloudflare argues that remotely hosted AI agents need protocols for Sites to identify agents and operators, establish appropriate trust in the Users they represent, and assign accountability without requiring agents to behave like humans.
- Agents obscure which Users are behind requests, making it harder for Sites to manage fraud, account sharing, compliance, resource use, and content monetization.
- CAPTCHAs only show that a human is present at a moment, while broad access scopes reveal neither the User nor who is responsible when an agent makes a mistake.
- Cloudflare favors Web Bot Auth for agents to sign requests, while noting that protocols must also represent chains of intermediaries and communicate other identifiers when needed.
- Identity can range from none to session-level, operator-account-level, or a unique per-User identifier; privacy-preserving approaches such as Privacy Pass and MOLE can establish trust without identifying Users.
- The proposed principle is to preserve the Site–User relationship and each participant’s security responsibilities, while allowing Sites and agents to agree on permitted behavior through protocols and standards.
#ai agents #web security #web bot auth #privacy
Generated by gpt-5.6-sol. Check the original for details.
Aliases
| Source | External ID | URL |
|---|---|---|
| Cloudflare Blog | 01M6HXVSQY6XGPSH12AFMB81B9 | https://blog.cloudflare.com/trusted-agent-delegation/ Read discussion |