OpenAI, Google, Microsoft and Anthropic are introducing a new standard that lets websites refuse the collection of data for AI training, The Verge reported on Oct. 31.

The report said website operators have so far had to tell each AI company separately whether they would allow their data to be used for AI training. Under the new standard, a single setting can apply to multiple companies at once.

The Verge reported that it was proposed to the Internet Engineering Task Force by Kevin Kelly (케빈 켈리), technical lead for Anthropic's standards work. It is a common signal allowing websites to notify multiple AI companies at once of their intent to refuse AI training.

OpenAI, Google, Microsoft and Anthropic said they will block access by their AI crawlers when they detect the token. The timing will differ by company. Anthropic is already following the standard. Google and Microsoft plan to apply it from next year, and OpenAI plans to apply it within a few months.

Andy Stone (앤디 스톤), Meta's head of communications, also voiced support but did not say whether Meta will follow it.

Microsoft is also offering detailed settings that allow users to separately refuse AI grounding, a process in which AI models refer to verified sources to reduce AI training issues and hallucinations.

Publishers have responded in various ways to the growing collection by AI crawlers. An RSL standard has emerged that charges crawlers a fee, and Cloudflare is blocking AI crawlers by default.