More than forty publishers, including the New York Times, Agence France-Presse, The Economist and The Atlantic, join the Standards for Publisher Usage Rights coalition. SPUR now brings together more than 80 members in more than twenty countries to establish common rules and tools to monitor the use of content by artificial intelligence systems. The initiative does not technically block training on its own, but rather seeks to make its use measurable, controllable and negotiable.
From dispersed trials to a common infrastructure
Editors have been questioning the aspiration of AI labs for their articles for several years. Lawsuits may seek compensation or clarify the law, but they proceed on a case-by-case basis, often over years. A common standard addresses another problem: before charging or authorizing use, it is necessary to know what content was consulted, for what purpose and through what system.
SPUR launched in February with founding members including the BBC, Associated Press and the Financial Times. Its expansion changes its potential weight. A convention adopted by some media may be ignored; A format used by dozens of international groups may become a business requirement for trackers, agents and model providers.
The coalition is funded by its members and is neither a court nor a regulator. Therefore, it does not decide whether the formation of a model falls under an exception to copyright in each country. Its role is to create the technical and contractual elements that allow rights holders to express their conditions, record usage and facilitate licensing agreements.
Telemetry should indicate what an agent actually consumed.
SPUR’s newly introduced content telemetry standard aims to standardize how an AI system records and reports its content usage. The problem goes beyond a simple robots.txt file, which tells robots whether they can visit an address, but doesn’t necessarily tell the story of what they extracted, selected, summarized, or used to generate a response.
Usable telemetry should identify the agent, the resource being accessed, the access time, and the usage category. This information can then be incorporated into an invoice, audit or compliance check. To work, the mechanism must remain simple enough to be adopted by many tools, while resisting incomplete or unverifiable statements.
The difficulty lies precisely in the incentive. An actor who accepts a license has an interest in providing reliable measurements; An operator who collects pages clandestinely can ignore the rule or hide his identity. SPUR will therefore have to supplement the specification with audit methods and contractual commitments. The coalition can improve the legal market without eliminating unauthorized uses.
Editors also defend the traffic that finances their newsrooms
Coaches’ compensation is only part of the conflict. Search engines and assistants increasingly answer questions directly, reducing the need to click to the source. Reuters cites a Reuters Institute study that found that Google organic search visits to more than 2,500 sites decreased by 33% between November 2024 and November 2025.
For a media outlet, this decline affects advertising, subscriptions and the discovery of its other content. A paid license for training data does not necessarily compensate for recurring loss of drives. Publishers therefore try to distinguish between several uses: corpus creation, real-time retrieval, summary in an interface, and citation accompanied by a link can have very different economic effects.
At the same time, judicial proceedings remain active. We had already mentioned the Encyclopaedia Britannica complaint against OpenAI and the agreement signed by Anthropic regarding pirated books. These cases concern different works and legal bases, but show why rights holders no longer want to rely on opaque bilateral negotiations.
A standard will not replace a balance of power
SPUR’s success will depend on corporate adoption of AI. If the major providers agree to a common format, publishers will be able to compare usages and negotiate on a more precise basis. If they refuse, the coalition will at least be able to document shared expectations and provide useful technical elements in contracts or disputes.
The opposite risk would be creating a system accessible only to large groups capable of managing complex licenses. Small media, blogs and independent authors also produce content used by generative systems. A truly sector-specific standard must offer clear options, low integration costs and intermediaries that do not capture the majority of the remuneration.
The new group now has more than 80 members spread across more than twenty countries, but it does not represent the entire world press. Its next test will be the publication of the operational terms of the Content Telemetry Standard: the reporting scheme, verification, error handling and compatibility with existing systems will determine whether it becomes a used infrastructure or a statement of principles.
