Cloudflare lets sites block AI training while staying search-visible
Cloudflare is giving site owners a way to remain discoverable while disallowing AI training, introducing new controls and an Accountable designation that establish a shared model with Apple, Google, and Microsoft for mixed-use AI crawlers. Publishers can now signal training refusal without sacrificing search visibility through Cloudflare's controls. Anthropic's separate threat intelligence report documents eight months of Claude abuse, illustrating why training-data boundaries matter. Chinese AI labs including Alibaba's Qwen team, DeepSeek, and Moonshot AI relayed requests en masse or extracted training data, with Qwen alone accounting for more than 151 million exchanges. Actors also used Claude for missile software and autonomous kamikaze drones. Site operators gain a vendor-backed split between discovery and training, while model providers confront API-scale distillation attempts.
Cloudflare lets sites block AI training while staying search-visible
Cloudflare is giving site owners a way to stay discoverable while disallowing AI training. New controls and an Accountable designation establish a shared model with Apple, Google, and Microsoft.
Key takeaway
Cloudflare's Accountable designation lets publishers block AI training crawlers while keeping content visible in search through a shared Apple-Google-Microsoft model.
What happened
Cloudflare announced new controls and an Accountable designation that let site owners stay discoverable in search while disallowing AI training, establishing a shared approach with Apple, Google, and Microsoft for mixed-use AI crawlers.
Anthropic's new threat intelligence report documents eight months of Claude abuse, with Chinese AI labs including Alibaba's Qwen team, DeepSeek, and Moonshot AI relaying requests en masse or extracting training data; Qwen alone accounted for more than 151 million exchanges, and actors also used Claude for missile software and autonomous kamikaze drones.
Evidence
Cloudflare introduced controls and an Accountable designation with Apple, Google, and Microsoft for mixed-use AI crawlers.
Cloudflare AI · attributed
New controls and an Accountable designation establish a shared model with Apple, Google, and Microsoft.
Site owners can stay discoverable while disallowing AI training.
Cloudflare AI · attributed
Cloudflare is giving site owners a way to stay discoverable while disallowing AI training.
Anthropic documented eight months of Claude abuse in a threat intelligence report.
The Decoder · attributed
Anthropic's new threat intelligence report documents eight months of Claude abuse.
Qwen alone accounted for more than 151 million exchanges tied to training-data extraction.
The Decoder · attributed
with Qwen alone accounting for more than 151 million exchanges
Actors used Claude for missile software and autonomous kamikaze drones.
The Decoder · attributed
Actors also used Claude for missile software, autonomous kamikaze drones
Why it matters
Publishers and API operators now face parallel defenses against industrial-scale training extraction, spanning web crawling policy at the edge and API abuse documented against frontier models.
Limits and uncertainties
The Decoder analysis notes Anthropic's report lacks verification of whether extracted data was successfully used to improve competitor models.
Practical implications
Builders and researchers should treat frontier model APIs as targets for data distillation and enforce rate limiting, anomaly detection, and strict usage policies.
Site operators on Cloudflare can adopt Accountable controls to separate search discoverability from AI training permission without a binary opt-out.
What to watch
Whether Apple, Google, and Microsoft crawlers consistently honor Cloudflare's Accountable designation in production traffic.
Whether Anthropic's countermeasures reduce ongoing large-scale extraction attempts such as the 151 million Qwen-linked exchanges.
