VTV.vn - OpenAI, Google, Microsoft and Anthropic commit to applying a common protocol, helping publishers control the use of content to train AI.

Photo: VTV

OpenAI, Google, Microsoft and Anthropic are committed to providing a unified way for news organizations and others to prevent their content from being collected for artificial intelligence (AI) training. The initiative aims to simplify data collection blocking, in the context of increased criticism and lawsuits from publishers against AI companies for using content without permission.

Currently, press agencies must set up separate blocking orders for each tool that automatically scans and collects data online. With the new solution, they only need to use a single standardized instruction. AI companies that apply this protocol will recognize and comply with that instruction.

However, it will be several months before the solution is fully implemented. Microsoft plans to apply from February 1, 2027, while Google will deploy in 2027. OpenAI, ChatGPT's developer, said it will apply in the coming months, while Anthropic, Claude's developer, has already begun deployment.

According to Microsoft's blog post, the new setting only applies to data collection for future model training, and does not prevent data collection for other purposes. Content remains crawlable by default until the publisher changes the setting.

In separate posts, Microsoft and Google also said that refusing to allow content to be used to train models could reduce the ability of publishers' content to appear in AI apps and experiences.

The announcement comes as publishers face a surge in tools that automatically scrape articles to feed data to AI models. Some press agencies have taken the dispute to court.

In December 2023, the New York Times sued OpenAI and Microsoft, accusing the two companies of using millions of articles to train ChatGPT without permission or payment. The two companies cited an exception in copyright law, known as fair use.

Content collection controls are also being implemented. From July 2025, Internet infrastructure provider Cloudflare will block AI data collection tools on new customers' websites by default. A standard called Really Simple Licensing also allows publishers to charge fees for tools that access their sites.

Mr. Andy Stone, Meta spokesman, expressed support for the new initiative on social network X. According to him, creating a common signal for publishers to communicate their wishes to many providers at the same time is an important step. Meta looks forward to collaborating with other AI companies to develop this standard and drive industry-wide adoption.