Ziff Davis CEO Pushes Back on DOJ AI Licensing Fee Claim

Better Measurement for AI Content Licensing, Not More Hype Publishers are arguing that AI economics should include sustainable licensing arrangements for the content used to build and operate generative systems. Capability is becoming abundant. Advantage will increasingly come from how well organisations combine technology with data, people, governance and customer experience. Teams should treat AI products as living operating systems, with continuous evaluation, feedback and clear ownership for the outcomes they influence. #Content #Licensing #Ziff #Davis #CEO #OpenAI #Find #750 #AI #ArtificialIntelligence #EnterpriseAI #MachineLearning #AIAdoption #ResponsibleAI #AIGovernance #AILeadership #AIStrategy #GenerativeAI #AgenticAI #AIAgents #TrustworthyAI #AISafety #HumanCentredAI #AIInnovation #AITransformation #AIProductivity #AIInfrastructure #FutureOfAI #EmergingTechnology #Technology #Innovation #DigitalTransformation #Leadership #BusinessStrategy #TechLeadership #FutureOfWork #DigitalStrategy #OperationalExcellence #Productivity #Transformation SOURCE NOTE Ziff Davis CEO: if OpenAI can find $750 billion for data centers, it can find money for publishers, too

AI companies need content from publishers, and they need to start paying. If not, we're going to end up with all the good stuff paywalled -- and AI tools pulling answers exclusively from marketing copy and astroturfed social media posts. That's a bad future for everyone. Thankfully, AI companies appear capable of raising unimaginable amounts of money, so it shouldn't be too difficult for them. https://lnkd.in/gTfzrnWH

OpenAI and Microsoft Knew They Were Starting a “Doom Loop” for the Web Recently unsealed court documents from The New York Times’ legal case against OpenAI and Microsoft have brought renewed attention to the broader impact of AI training practices on the internet. According to the documents, internal discussions warned that large-scale data scraping could contribute to a “doom loop” for the web. The concerns centered on how AI-generated answers might reduce traffic to publishers, weaken online business models, and ultimately limit the availability of high-quality information for future AI systems. The documents also reportedly described the use of online content to train AI models in highly critical terms, including references to the “largest theft of labor in human history” and claims that the practice made “a complete mockery of the idea of fair use.” Several of the most notable statements were attributed to Microsoft Director of Applied Science Brent Hecht. Microsoft has sought to distance itself from those comments, emphasizing that they represented individual views rather than official company policy. The case highlights unresolved questions about copyright, consent, attribution, publisher sustainability, and the responsibilities of companies developing artificial intelligence. As the legal process continues, the debate over how AI systems access and use human-created content is likely to shape the future of the web. Artificial Intelligence School offers expert-led programs designed to help professionals understand and apply artificial intelligence responsibly and effectively. #aischool #artificialintelligenceschool #ArtificialIntelligence #AI #GenerativeAI #Copyright #DigitalMedia #Technology #FutureOfWork #ResponsibleAI

The Verge’s framing of a looming “doom loop” is dead serious: AI summaries reduce clicks, publishers lose revenue, content quality drops, and the training pool gets worse. OpenAI and Microsoft allegedly knew this dynamic-and it’s exactly what happens when intermediaries capture value while creators eat the cost. This isn’t just a media problem. It’s an enterprise knowledge problem too. If your business relies on other people’s content, and you’re building AI products that strip attribution and traffic, you’re quietly sabotaging the ecosystem you depend on. My blunt view: AI companies should be forced to pay a “content exhaust tax” - per query, per domain, transparently - or be legally blocked from summarizing the web at scale. If that sounds extreme, ask yourself why creators should fund the intelligence layer for free while platforms monetize the output. Best practice for leaders: treat data provenance and attribution like security - design it in, audit it, and pay for what you use. If AI makes the open web economically impossible, what exactly will it learn from next year? #GenAI #DigitalMedia #AIethics #OperatorOS Playing it forward (get my CEO Playbooks): https://lnkd.in/gr_x7hgi https://lnkd.in/g7qxsvR7

Google, Microsoft and OpenAI have all signed the EU's code of practice on AI-generated content. About half of all the signatories are small and recent companies. That's the European Commission's own description. About 190 organisations had signed by the end of July, and the list has kept growing. It's easy to file the EU AI Act under "big tech's problem". The list says otherwise. As of 17 September it shows 95 signatures on the section for companies that build AI tools that generate content, and 192 on the section for businesses that use those tools and label what they publish. Some have signed both. Two things to know before anyone tells you to sign it. Signing is voluntary. The duties underneath it aren't. It's also not a one-off. The Commission is launching two task forces for signatories this month, to share practice and feed back on how the rules are working. The full list, from the Commission: https://lnkd.in/ewx6Cexq

Microsoft's own scientists called AI scraping "the largest theft of labor in human history." Newly unsealed filings in the New York Times copyright suit against OpenAI and Microsoft reveal what else was said in private. Microsoft's own data showed its Copilot answer engine cut click-throughs to the Times' domain by as much as 93% versus traditional Bing search. A January 2024 presentation by Brent Hecht, a Microsoft director of applied science, called it a "doom loop" that would hurt both the models and the open web. OpenAI's head of ChatGPT, Nick Turley, wrote that publishers face an "existential threat" from products that are "largely substitutive" and "will get more and more substitutive as they get better." Satya Nadella testified that paywalled content should be licensed for training. The scale is the other half. OpenAI's mid-training datasets held more than 91,692 copies of works from the Times, Daily News and the Center for Investigative Reporting. A Common Crawl dataset carried more than two million nytimes.com documents. One caveat that matters: most of this comes from the Times' own brief, and the underlying exhibits remain sealed. If your product answers questions using someone else's reporting, what is your plan for when the fair use argument stops being free? #AI #Copyright #Media #TechPolicy #OpenAI

🔊 An AI-hungry Google isn't playing nice with one of the largest content publishers on the internet. "Here's the truth. For a long time, we had a deal with Google, an unwritten deal, which was 'you can use our content to build your search product, and in return, we get sessions,'" People Inc. CEO Neil Vogel told me from the Goldman Sachs Communacopia & Tech Conference. "Now the search product has turned into an AI product by and large, and they still use our content to train their AI, both in real time and in a training way — we don't have to get into the AI terminology of it — but now they don't share anything with us." Google's push into AI-powered summaries has upended search traffic for publishers. Many fear the situation will get worse, triggering more sweeping cost cuts among many players in 2026. Vogel said the company has signed content licensing deals with Microsoft, Meta, and OpenAI. Ideally, he would also like to ink one with Google. Deeper dive: https://lnkd.in/g7edfNcH

Anthropic, OpenAI and Google have been holding private working-group sessions since July to build their own shared industry standards body. The Information reported it on September 13. The same three labs are among the companies required to file their first EU AI Act systemic risk evaluations by today, September 15. That filing is not light. Red-teaming methodology, energy consumption, a completed copyright training summary on a template the AI Office published in July. Independent third-party evaluators, not the labs themselves, must sign off on the adversarial testing. Non-compliance can reach 3% of global annual turnover. No fine has been issued under that provision yet. If the industry body is recognised under the EU AI Act's harmonised standards process, its protocols could satisfy parts of those same technical requirements. The companies filing the form today are also assembling the group that may write what the form asks for next. And under the Act, deployers carry obligations of their own regardless of what the provider files. Which of your vendors has told you in writing whether their flagship model carries a systemic risk designation under Article 51? https://lnkd.in/dKP3i8hM

AI News Roundup — September 2, 2026 A few stories worth your attention if you're building with AI right now: Compute keeps compounding Anthropic signed a $35B, six-year cloud deal with Nvidia-backed Lambda, pushing its 2026 compute commitments past $135B. If you've noticed frontier labs announcing infrastructure deals more often than new models lately, this is why — training and serving at scale now runs on multi-year power and chip commitments, not spot capacity. Regulation catches up to chat interfaces The EU designated ChatGPT a "Very Large Online Search Engine" under the DSA, triggering systemic risk assessments and audits by late November. The logic: if your assistant searches the web and answers billions of queries, regulators will treat it like a search engine no matter what you call it. Worth watching if you're shipping agents with browsing built in — the compliance bar just moved. Training data lawsuits escalate Sony and Warner sued Anthropic, alleging around 20,000 songs were used to train Claude via pirated sources, with potential damages in the billions. It's the clearest sign yet that music rights holders are following the same legal playbook as publishers and authors before them — data provenance is now a first-class engineering concern, not a legal footnote. Agents are getting practical guardrails GitHub published a production case study on "PR Sous Chef," a Copilot workflow that checks open PRs every 15 minutes and decides when a human actually needs to step in. Separately, SonarSource shipped "Sonar Vortex," a code graph built to cut the token cost of agents that otherwise burn budget on repeated greps and full-file reads. Same trend, two angles: 2026's agent work is less about raw capability, more about making agents cheap and reliable enough to run continuously. What's catching your eye more right now — the compute race, the regulation, or the copyright fights?

The difference between established search engines like Google and emerging LLMs lies in data and experience. Google has decades of search data, billions in revenue, and has navigated countless legal challenges, all of which inform and refine its algorithms. This deep well of experience and data allows for a nuanced understanding of information and user intent that LLMs are still striving to achieve. As LLMs evolve, they will inevitably face similar issues, learning through experience and potential legal precedents, much like Google has. #AI #LLMs #Tech #Innovation #SearchEngine