Byteful, the Manchester-based proxy provider, has started blocking traffic to the AI inference APIs of OpenAI (api.openai.com) and Anthropic (api.anthropic.com) on its network by default as of 30 October 2026. The provider cites a higher risk of abuse on this kind of traffic, such as foreign AI distillation attacks, and says more than 99.9% of customer traffic is not affected.
According to Byteful's product update log, customers with a legitimate business reason can request access — but only if they are based outside Hong Kong and China. Requests are reviewed case by case and require KYC verification; access requests had to be submitted before 30 October to be in place when the block took effect.
Context: direct IP access already restricted
The AI API block is the second recent default restriction in Byteful's network. Requests that target raw IP addresses instead of hostnames have been blocked by default since 7 September 2026, also with case-by-case exceptions. Beyond these two, Byteful maintains a private blocklist of over 6,000 items covering government domains, banking and payment processors, NSFW sites and mail services, and enforces Internet Watch Foundation blocklists as an IWF member.
For users of Byteful's residential, ISP, datacenter and mobile proxies, the practical effect is that scraping or routing requests through api.openai.com or api.anthropic.com will fail unless an exception has been granted, and exceptions are not available to customers in Hong Kong or China at all.