The summer the
compensation rulebook moved
As AI tools are developed and deployed at speed, new market mechanisms for licensing and compensating content are taking shape, with companies from large corporations to independent start-ups building ways to connect AI firms with the content they rely on. As Courtney Radsch and Karina Montoya observe in their report Same Gatekeepers, New Tollbooths, all of this is happening in a public-policy and regulatory vacuum, producing a range of risks and market externalities. Collective and independent efforts are both under way, but powerful Big Tech firms are once again trying to set terms that work mainly for them, and to capture the market.
The AI licensing and compensation tech stack is being built
The technical stack that aims to let AI companies and publishers decide who can access content, how it can be used and who should pay for it moved this summer toward machine-readable licensing terms and standards. The SPUR publishing-industry coalition (around 30 publishers, including the BBC, the Financial Times, the Guardian, Sky, the Telegraph, Mediahuis and AP) published its first draft Content Telemetry Standard, which lets a publisher see whether their content was crawled, cited or used to ground an AI answer. Really Simple Licensing (RSL) lets publishers set machine-readable terms covering attribution, pay-per-crawl and pay-per-inference. And from 15 September 2026, Cloudflare blocks mixed-purpose AI crawlers by default on pages carrying advertising.
Content marketplaces are another piece of infrastructure taking shape, and they are being built by Google, OpenAI, Cloudflare and Amazon, the same companies whose crawlers created the problem in the first place. If the sector's entire strategy is to get a better price inside their systems, we will end up exactly where we did with AdTech: dependent on infrastructure controlled by the same tech companies that cannibalise our content.
The licensing tools being built are, for now, tools for publishers with money and engineers. Small and community outlets are being left out by default.
Do thisHelp interested organisations and newsrooms switch on the one free protection that already exists, robots.txt, and push for small- and minority-language-publisher seats as the paid standards (SPUR, RSL) are written. Today only around 30% of South African news sites, and 7% in Brazil, block even one AI crawler.
More detail: the tech stack and the Protocol Gap evidence
The licensing and telemetry layer
- SPUR published its first draft Content Telemetry Standard in June. It lets a publisher see whether content was crawled, cited or used to ground an AI answer, while protecting both user privacy and intellectual property.
- Really Simple Licensing (RSL), developed by the non-profit RSL Collective, is an open standard for machine-readable terms covering attribution, pay-per-crawl and pay-per-inference. Rightsholders can choose between compensation models or prohibit AI use altogether.
- Cloudflare's default applies to new customers, new sites and free-plan users, and site owners can change it. By default, ad-carrying media content may no longer be used to train AI models.
The Protocol Gap evidence (IFPIM audits)
- IFPIM, working with GIBS (South Africa), Momentum (Brazil) and AMSI (Indonesia), audited the use of robots.txt and published the results as a series of Protocol Gap reports.
- South Africa: three-quarters of news sites carry robots.txt, but only around 30% use it to block even one AI crawler, and those same sites lost, on average, 20% of their daily page views in a year (The Protocol Gap: South Africa).
- Brazil: three-quarters again carry the file, but just 7% block any AI crawler (The Protocol Gap: Brazil). Findings from Indonesia are due to be published soon.
- In both countries it is the large, well-resourced publishers doing the blocking.
Competition investigations and copyright rulings are now the more promising route
Over the past months, the compensation and licensing question has moved out of private, largely bilateral licensing deals with the biggest news companies and into the public-policy and regulatory arena, where the leverage is stronger and the outcomes reach every publisher, not just the one at the table.
Regulators and courts are now a faster route to getting paid than negotiating with a platform alone.
Do thisIf a competition authority in your jurisdiction is examining platform-news relationships, submit your own revenue and visibility-drop data, prepare a court submission such as an amicus brief, or organise a collective meeting with the authorities. Anchor it in two precedents: Italy, where copyright can be reactivated to build a licensing model, and the UK, where opting out of AI Overviews must not cost you search ranking.
More detail: the six cases, jurisdiction by jurisdiction
- Italy / EU (CJEU): the Court ordered Meta to compensate Italian publishers in a case brought by media regulator AGCOM. It confirms that Member States can set up fair-payment mechanisms for the online use of press content under Art.15 of the 2019 Copyright Directive, demands transparency in defining fair remuneration, and says platforms must negotiate for the commercial use of journalistic content. The judgment strengthens the position of other Member States facing similar challenges.
- France: nearly 300 newspapers filed a complaint with the Autorité de la concurrence over Google's AI Overviews, launched in France in late July. Publishers report 33 to 38% traffic losses. APIG argues that AI Overviews breach Google's 2022 copyright commitments, the same commitments that already cost Google €500 million in 2021 and €250 million in 2024 in fines.
- Nigeria: President Tinubu directed the FCCPC to investigate Meta, Google, X and generative-AI platforms over alleged anti-competitive practices and unauthorised use of Nigerian news content, following a joint petition from the Nigerian Press Organisation.
- Brazil: competition authority CADE issued a 180-page opinion recommending formal antitrust proceedings against Google, framing AI Overviews as an "exploitative abuse of dominant position" and reframing the issue from narrow scraping to a broader market-structure question.
- EU: in December 2025 the Commission opened a formal Art.102 TFEU investigation into Google. In February 2026 members of the European Publishers Council (among them DMG Media, the Guardian, News UK and the New York Times) filed a complementary complaint. The Commission sent publishers a summer questionnaire on whether Google's opt-out is viable without costing search visibility.
- UK: the CMA ordered Google to let publishers opt out of AI Overviews, AI Mode and broader generative-AI services, at both directory and page level, without harming ordinary search ranking, and to credit any content it uses with a direct link. The order follows a June 2025 complaint from Foxglove documenting traffic drops of up to 79% for some UK publishers.
UNESCO's Draft Guidance on Fair Compensation for News
Intended as a reference model for national regulatory frameworks, it is built on three pillars (rights-based, competition-based and fiscal), combinable through four disbursement models, under nine guiding principles (para. 45). By the 30 July deadline more than a dozen journalism and media submissions were on the table. GFMD analysed six in full, plus the CCIA counter-position.
News compensation and AI compensation are not the same problem. Separate press-publisher rights from AI governance, letting publishers permit search indexing while refusing AI training or summaries.
The definition of media is too narrow. Cover the full range of journalistic output (TV, radio, podcasts, scripts, subtitles, archives and multilingual services), with explicit accommodation for diaspora and exile media.
Smaller media need dedicated protection. Separate funding streams and eligibility rules for regional, non-English outlets, plus a protection-and-insurance mechanism for those who cannot afford to litigate.
Pay journalists as individual rightsholders. Training deals pay corporate rightsholders, not the individuals whose labour produced the work; follow EU equitable-remuneration rights that survive assignment.
Enforcement needs to be structural. Enforceable obligations with deadlines and sanctions, including a "must-offer" right and a banking or pharma-style licence for very large platforms.
The technical layer needs governance too. Non-discrimination rules for crawlers, standardised transparency, and attribution and compensation at the model level, not just the product.
The CCIA (Computer & Communications Industry Association) is a US tech-industry lobby group whose members include Google, Amazon, Meta and Apple. In its UNESCO submission it rates the three-pillar architecture "not very accurate or useful," would put voluntary, market-based arrangements on equal footing with the interventionist pillars, and proposes adding four principles: free flow of information, competitive neutrality, trade and copyright consistency, and evidence-based policymaking. It calls the "siphon" premise empirically unsupported and frames AI training as transformative fair use. It shares one point with civil society, that the Draft is thin on implementation detail, but reads that as grounds for less intervention.
UNESCO's final text is being drafted now, and all six community asks fit inside the structure it already uses, which makes them realistic to win.
Do thisUse the same six points as ready-made arguments in national competition cases, copyright reform and AI-governance consultations, anywhere a framework for paying for journalism is being written.
More detail: the architecture, the nine principles, who submitted
Three pillars and four disbursement models
- Rights-based (neighbouring rights, mandatory licensing), competition-based (mandatory bargaining, structural remedies) and fiscal (levies, subsidies, tax credits).
- Combinable through four disbursement models: collective management (for example Denmark's DPCMO), direct negotiated payments, an industry-led trust fund, or a public fund (for example the Canadian Journalism Collective).
- The Value of Journalism companion brief supplies the evidence that compensation is a rational public investment, not charity. Final guidance and a summary of contributions are due later in 2026, followed by three regional roundtables.
The nine guiding principles (para. 45)
- Human rights and freedom of expression · journalism and news as a public good · equity, inclusion and media plurality · transparency and accountability · independent oversight · attribution and recognition · fairness and economic balance · technology adaptability and future-readiness · multi-stakeholder participation.
Submissions analysed
- AJOR; the Center for Media and Digital Governance at the Open Markets Institute; RNW Media; the Association for International Broadcasting; and a joint response from Denmark's DPCMO with Anya Shiffrin (Columbia University) and Haaris Mateen (University of Houston). As the platform-industry counter-position, GFMD also analysed the CCIA submission.
Article 50 and the provenance layer, built without small publishers
The EU AI Act's Article 50 (in force 2 August 2026) requires AI providers to label synthetic content, and obliges deployers, newsrooms included, to disclose AI-generated and manipulated content. Fines run to €15m or 3% of global turnover. Alongside it, a provenance stack (C2PA content credentials, IPTC's Origin Verified Publisher Certificate) lets publishers prove what is authentic. But adoption is concentrated among large publishers with the tools and verified identity to opt in. The very mechanism that will decide what counts as authentic is being built without the small publishers who most need to be verifiable inside it.
A new layer that decides which content counts as "authentic" is being built, again mostly by and for large publishers.
Do thisGet members into the Code of Practice's Working Groups and Signatory Taskforces now, while the rules for how labelling works are still open (only around 10 newsrooms have signed so far), so small and minority-language outlets are verifiable inside the system rather than shut out of it.
More detail: how Article 50 works, and the provenance tools
Article 50 in practice
- Providers must mark their outputs as AI-generated. Deployers (newsrooms, CSOs) must disclose AI-generated content to their audience, including deepfakes and public-interest text. No special software is needed, as deployers can label content with the EU's official icons.
- Exemption: the duty does not apply once content has had genuine human review, such as fact-checking and a named editor or organisation taking formal responsibility. The exemption falls away the moment anything is substantively changed afterward.
- It also targets the "liar's dividend," the way the mere existence of deepfakes lets bad actors dismiss genuine reporting as "probably fake too."
- Backed by official Guidelines and a voluntary Code of Practice with around 230 signatories; the Signatory Taskforces launched in September 2026.
The provenance stack
- C2PA (Content Credentials): cryptographically signed origin and edit history, already piloted by major newsrooms, but it depends on cameras, editing tools and content-management systems.
- IPTC Origin Verified Publisher Certificate: publisher-identity attestation that vouches for who published content, but it assumes identity verification. Each step is manageable for a large media company and a real cost for a small newsroom.
- The Journalism Trust Initiative (JTI): an international, ISO-type standard initiated by Reporters Without Borders (RSF), designed to promote trustworthy news media and protect content provenance. It evaluates editorial integrity and operational transparency across 130 specific criteria.
28 processes,
two questions
The tracker maps publisher remuneration and compensation frameworks, tools, and processes. Each is assessed from two perspectives: its policy significance (relevance to members, bindingness, and reach) and its actionability for the GFMD community (whether a window is open, and whether it is a priority for the public interest information community).
The register
Click any row for the stage, the latest status note and the primary source. Sort by any column heading.
How the scores work
Policy significance is the mean of relevance to members, bindingness and geographic reach: how much a development matters on its own terms. GFMD actionability is the mean of process stage and positional advantage: whether a window is open and whether this is GFMD's lane. The quadrant threshold is 3.5 on each axis.
Open for
contribution
UN Special Rapporteur — 2027 vision & priorities
The new mandate invites views to inform its 2026–2029 priorities. Submissions up to 2,500 words (EN/FR/ES), deadline 31 Oct. Add evidence and regional examples to GFMD's joint response by 19 Oct — lbecanaball@gfmd.info.
European Commission — consultation on technology & copyright
A targeted EU consultation (published 29 Sep) is open until 3 November 2026, covering copyright-protected content in AI, online piracy of live events, the single equitable remuneration right, and copyright of scientific research. Press publishers are not named, which is the opening to put journalism and news into the AI-and-copyright section. Respond via the EU Survey questionnaire. (Tracker C28.)
France — Autorité de la concurrence (APIG complaint)
APIG's complaint was filed 11 Aug; an opening decision is awaited (the French administrative-court cadence is typically two to six weeks). Route in: third-party and amicus-style submissions from independent French publishers outside APIG and from European small-publisher coalitions. The Autorité settles the shape early, so naming distributional concerns now matters. (Tracker C18.)
UNESCO — Fair Compensation for News, final guidance
Consultation closed 30 July; UNESCO is drafting now, with three regional roundtables to follow. Shape it through direct outreach to the Communication & Information Sector and the roundtables once announced. (Tracker C01.)
WIPO SCCR/49 — Standing Committee on Copyright and Related Rights
The 49th session runs 30 November – 4 December 2026 in Geneva. Copyright and generative AI, broadcasting and audiovisual works are on the provisional agenda; a dedicated journalism or cross-border news-royalty item is not. With broadcasting and audiovisual organisations already represented, there may be room to add other information providers. (Tracker C20.)
IGF 2026 — "Governing the Internet in the Age of Intelligence"
Main sessions finalised by 19 Oct; registration open; bilateral meetings with delegations can be arranged from 9 Nov. Contribute via the Dynamic Coalitions on the Sustainability of Journalism, on Data & AI Governance, and on Community Connectivity.
Geneva AI Summit 2027 — minimum standards on extreme AI risks
The Swiss National Council (2 Oct 2026) asked the Federal Council to use the summit to negotiate minimum safety standards for the most powerful AI models; ICT4Peace is pressing for matching ambition and resourcing. The focus is extreme risks (biological, chemical, cyber, loss of control), adjacent to compensation but an early venue to shape how AI governance treats information.
Build inclusion in from the start
Design fair compensation frameworks, mechanisms, standards and tools to reach and benefit small publishers deliberately. Inclusion has to be built in from the start: low entry thresholds so small outlets qualify, caps so the largest publishers can't take most of the money, extra weighting and reserved funds for smaller and minority-language outlets, guaranteed seats where technical standards are written, and a clear definition of who counts as journalism.
Models exist: Canada shares Google's payments by journalist headcount and caps broadcasters' share; Australia raised its loading for regional, small and diverse publishers to 20%; and South Africa's Digital News Transformation Fund is reserved for small and community media. Without such design, small and minority-language outlets are excluded by default.
Reactivate copyright against AI
Reactivate copyright and press-publishers' rights frameworks against AI. France's APIG complaint of 11 August 2026 is the first tracked case doing this. Every jurisdiction with a comparable regime, including Germany, Spain, Italy and Poland, should examine whether its own framework already covers the harms generative AI produces and act accordingly.
Treat publisher-identity attestation as a democratic safeguard
The debate around Article 50 should not collapse into a narrow technical argument about labelling synthetic content. With the EU AI Act's Article 50, the Code of Practice on Transparency of AI-Generated Content, and emerging provenance standards such as C2PA and IPTC's Origin Verified Publisher Certificate, small publishers, community media and minority-language outlets need seats inside the standards bodies, not just compliance obligations.
Fund journalism as public-interest civic infrastructure
Fund, safeguard and verify journalism as public-interest civic infrastructure, and treat journalists as first responders. Journalism is critical during crises, elections and disinformation events. Compensation, funding and provenance should be understood as three registers of the same public-good argument rather than three separate conversations.
The evidence base is set out in the Value of Journalism brief and echoed in the South African Competition Commission's remedies, which fund newsroom innovation, community media, and training for outlets publishing in local African languages.
Submit your evidence where authorities are investigating
If a competition authority in your jurisdiction is examining platform-news relationships, it's now the right time to submit the data and records of your revenue and visibility drop. Prepare a court submission (for example an amicus brief), or organise a collective meeting with the authorities where you and other media can present the evidence and strengthen the case.
Copyright frameworks can be reactivated to build in a licensing model, as in Italy; and the UK's opt-out-without-penalty precedent is exportable to other dominance cases.
Engage the standards Working Groups
Only around ten news publishers have signed the Code of Practice so far, a gap or an opening depending on how quickly the sector moves. The Working Groups that developed the Code will keep being active; if your organisation and newsroom wants to engage in developing this and other standards, as AI labelling is just one of them, finding ways to engage with the Working Groups is the right thing to do.
"Inclusion cannot be secured by general commitment alone; it depends on identifiable, verifiable design choices."
Suggested phrasing for submissions: meaningful inclusion of small, community and minority-language media requires deliberate design, not passive assumption — the compensation, provenance and infrastructure frameworks now under development share the same levers: eligibility criteria, funding allocation, and representation in standard-setting bodies.
Primary sources and reporting behind the briefing and the tracker, grouped by theme. Each tracker row also carries its own primary-source link.
The compensation tech stack
- Same Gatekeepers, New Tollbooths — Radsch & Montoya, Open Markets Institute
- SPUR Content Telemetry Standard — Digiday
- Really Simple Licensing (RSL) — rslstandard.org
- Cloudflare and the agentic web — Cloudflare
- The Protocol Gap: South Africa — IFPIM (PDF)
- The Protocol Gap: Brazil — IFPIM (PDF)
Competition & copyright
- Meta v. AGCOM, CJEU — Council of Europe / IRIS Merlin
- France: APIG complaint to the Autorité — Le Monde
- Nigeria: FCCPC investigation — FCCPC
- Brazil: CADE opinion on Google — CADE
- EU: Art.102 probe & EPC complaint — European Publishers Council
- UK: CMA opt-out order — Foxglove
Provenance & AI transparency
- EU AI Act, Article 50 — artificialintelligenceact.eu
- Code of Practice on AI-generated content — European Commission
- C2PA / Content Credentials — c2pa.org
- IPTC Origin Verified Publisher Certificate — iptc.org
- Global Media Identifier (GMI) for distribution channels and brands — ISO
- US AI-training copyright litigation — Reuters
Standards, multilateral & engagement
- UNESCO Guidance on Fair Compensation for News — UNESCO
- UNESCO, The Value of Journalism — UNESDOC
- CCIA comments on the UNESCO guidance — ccianet.org
- WIPO SCCR/49 provisional agenda — WIPO (PDF)
- EU consultation on technology & copyright — European Commission
- Information integrity as a UK G20 priority — Media20
- Strengthening media's bargaining power over generative AI — DataTank
- Geneva AI Summit 2027 — ICT4Peace
- OSCE policy manual: safeguarding media freedom in the age of Big Tech and AI — Forum on Information & Democracy
- Safeguarding access to reliable information in the age of AI (background report) — Forum on Information & Democracy (PDF)
GFMD Monthly Intelligence Brief, Issue 2026-10. Prepared by the GFMD Policy Team as a pre-read for the compensation & provenance policy meeting on 13 October 2026.
Contact: communications@gfmd.info
Agenda
Compensation Frameworks for News, a meeting in the GFMD Policy & Advocacy series. Most compensation and provenance systems are being built without journalists in the room; this session maps the processes that matter and the openings for collective action.
Welcome and introduction
Opening remarks from the moderator (Global Forum for Media Development).
Main processes mapped
Brazil's CADE investigation into Google's use of news content, the UNESCO Guidance on fair compensation for news, and the wider global trends and national cases.
Details Brazil's investigation, how other organisations are replicating it, and how to engage.
Covers Brazil's investigation and how regional efforts can connect to global initiatives such as CTRL+J.
Updates on South Africa's processes and their influence on other models.
Updates on the next steps for the UNESCO Guidance on fair compensation for news.
Provides an overview of global trends shaping fair compensation initiatives, the emerging AI content licensing market, and opportunities for collective action, based on her recent report and white paper, as well as recent updates from the Global Tracker on tech & media fair compensation frameworks.
Other related processes and advocacy opportunities
An open floor for members to share related cases and advocacy opportunities.
Presents the workstream on safeguarding access to reliable information in the age of AI, its background report, and progress on the recommendations on AI and media viability and visibility.
What is needed to influence these processes?
Next steps for collective action on the compensation processes mapped in the session.
Any other business
Updates on upcoming meetings, deadlines and other issues raised by members.
Closing
Closing remarks from the moderator.
To contribute a case or to speak, contact Ivana Bjelic Vucinic (ivbjelic@gfmd.info) and Laura Becana Ball (lbecanaball@gfmd.info) with a short summary, the background, and lessons that transfer to other countries or regions.
GFMD Monthly Intelligence Brief, Issue 2026-10. Prepared by the GFMD Policy Team as a pre-read for the compensation & provenance policy meeting on 13 October 2026.
Contact: communications@gfmd.info