AI agents are a ‘bane’ for some website operators, says Cloudflare
Reported by The Register and news.lavx.hu2 sources ↓
Cloudflare CEO Matthew Prince said some customers want to block AI agents because they can drive up website costs without reliably bringing business. He also said attackers may have exploited PHP vulnerabilities before disclosure and argued container-based systems may struggle with agent traffic.
Key points
- Agents can incur costs for sites they visit without generating business for most of them
- Cloudflare expects agent traffic could reach 1,000 times human click traffic
- Prince said about 5% of the web opted into Cloudflare content protection products since July 1
- Prince said Cloudflare is checking whether attackers exploited PHP vulnerabilities before disclosure
Sources
Details
On Cloudflare’s third-quarter earnings call, CEO Matthew Prince said some customers consider AI agents harmful to their businesses and want to block them. He described an agent visiting 1,000 restaurant websites to compare menus, while its user might choose only one restaurant, leaving the others with costs but no benefit.
Prince said Cloudflare expects agent traffic could eventually be 1,000 times human click traffic. He argued that container-based systems used by hyperscalers and newer cloud providers may not handle that traffic cost-effectively, and promoted Cloudflare’s serverless Workers service as an alternative. The Register noted that Cloudflare also sells products to block, manage or monetize AI-generated traffic.
Prince said about 5% of the web had opted into Cloudflare’s content protection products since July 1, and named Meta as a customer. He also said Cloudflare was reviewing its logs for evidence that attackers, including nation-state actors, used PHP vulnerabilities before they were disclosed; he expected the review to find examples.
Written from coverage as of
Images
- The Register
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh