Regulating AI-driven content discovery

Yesterday, the open internet got its first toll booth. Not from courts or Congress, but from Cloudflare. The company, whose infrastructure touches roughly 20% of all global internet traffic, announced a significant shift: All new customer domains will now block AI bots by default. That means crawlers like OpenAI's GPTBot, Anthropic's ClaudeBot, and Google's Extended bot can no longer freely help themselves to your content. Unless you, the website owner, explicitly allow it, they’re cut off. Cloudflare also announced a Pay‑Per‑Crawl marketplace, where publishers can set terms for access - turning content into a licensed input, not a free good. With one product update, Cloudflare rewrote the default terms of engagement between AI and the web. From open to closed. From assumed permission to enforced consent. From passive scraping to active negotiation. Cloudflare has forced the question that’s been dodged for years: If AI systems are built on the backs of human expression, who owns the value they create? ▪️ For 20+ years, the internet operated on a basic exchange: Publish freely → Get discovered via search → Monetize attention. But AI broke that contract. Users now ask Claude or ChatGPT for answers. The models reply using human-created content - with no credit, no clickthrough, and no compensation. When publishing stops being a path to discovery - and starts being a donation to model training - incentives collapse. What Cloudflare did was reintroduce friction. Not to break the web. But to rebalance it. ▪️ What makes this moment so interesting is that the solution isn’t coming from regulators. It’s not the result of a moral awakening or a philosophical reckoning. It’s a default setting in a SaaS dashboard. But because of Cloudflare’s scale, it might as well be policy. This is capitalism at its best. You don’t always need top-down regulation to enforce guardrails. Sometimes market incentives create their own form of governance. In a world where governments often lag behind technology, infrastructure becomes policy. ▪️ Cloudflare may not have intended to reshape digital economics, but they’ve done exactly that. By forcing AI firms to ask, license, or walk away, Cloudflare has built the bones of a content licensing infrastructure, without ever saying the word “copyright”. ▪️ Let’s be clear: Cloudflare didn’t do this because it’s noble. They did it because it’s profitable. Their customers - TIME, Condé Nast, Reddit, News Corp - are tired of being strip-mined for training data without so much as a backlink. And Cloudflare is in the perfect position to intervene. I won’t be surprised if Fastly, Akamai, maybe even AWS follow. For 30 years, we assumed the internet had no gatekeepers. But it did. We just didn’t notice them until they started saying no. The principle is planted: Scraping without consent is not neutral. It’s extraction. We could see a more cohesive consent layer for the web emerge - driven not by altruism, but by incentive alignment.

"The report outlines four key regulatory approaches to AI governance—industry self-governance, soft law, regulatory sandboxes, and hard law—each offering distinct advantages and challenges: 1. Industry Self-Governance • Strengths: Can directly impact AI practices if integrated into business models and company cultures. • Limitations: Non-binding; not appropriate for sectoral use-cases with particularly high risks – e.g. financial sector or healthcare; risk of ‘ethics-washing’. 2. Soft Law • Strengths: Soft law includes nonbinding international agreements, national AI principles, and technical standards, providing adaptable frameworks that promote responsible innovation. Early governance efforts by intergovernmental bodies have set important precedents. • Limitations: While soft law encourages innovation, it focuses on high-level principles rather than binding rights and responsibilities. 3. Hard Law • Strengths: Binding legal frameworks provide clear, enforceable guidelines that ensure AI stakeholders comply with established standards and regulations. • Limitations: Given the rapid pace of AI development, hard laws risk becoming outdated and can be extremely resource-intensive to implement. 4. Regulatory Sandboxes • Strengths: These controlled environments allow for real-world experimentation with AI technologies, supporting innovation and providing valuable insights without exposing the public to unchecked risks. • Limitations: Sandboxes can be resource-intensive and have limited scalability, making them less feasible for wide-scale governance across diverse sectors." Read/download: https://lnkd.in/etwyUaUK

The rapid rise of AI-generated media - particularly deepfakes and convincingly altered content - brings us to a crossroads in how we interact with information. Suddenly, seeing isn't necessarily believing. This shift raises critical questions: How do we verify what’s real, and how do we address creators' intentions behind such content? Do we just categorize them as creative output? Addressing this challenge likely requires multiple, coordinated approaches rather than a singular solution. One fundamental strategy involves enhancing public media literacy. Teaching ourselves and our communities how to recognize misinformation, critically evaluate sources to help reduce the spread of misleading information. Initiatives like educational campaigns, school programs, and public-service messaging could strengthen our collective defenses against misinformation. Simultaneously, technology companies producing or distributing AI-generated content could implement practical measures to build transparency and trust. For instance: - Clearly watermarking content generated by AI tools. - Requiring upfront disclosures about synthetic or substantially altered media. - Employing specialized authenticity verification technologies. Moreover, adopting clear ethical standards within industries utilizing AI-driven media - similar to those upheld in professional journalism - could encourage greater accountability. Finally, regulatory frameworks will be important - but they must be carefully designed. Excessive restrictions could inadvertently stifle innovation and legitimate expression. Conversely, too little oversight leaves society vulnerable to harmful deepfakes, especially in contexts like elections. Targeted and balanced regulations can minimize harms without impeding creative and productive uses of AI. Where should efforts be prioritized most urgently - strengthening public awareness, establishing clear industry standards, or developing nuanced regulatory policies? #innovation #technology #future #management #startups

The European Commission published its first draft of the “Code of Practice on Transparency of AI‑Generated Content” designed as a tool to help organizations demonstrate alignment with the transparency requirements (Art. 50) of the AI Act. Article 50 of the AI Act includes obligations for providers to mark AI-generated or manipulated content in a machine-readable format, and for users who deploy generative AI systems for professional purposes to clearly label deepfakes and AI-text publications on matters of public interest. The document is divided into two sections. The first section covers rules for marking and detecting AI content, applicable to providers of generative AI systems, including to: - Use a Multi‑layered machine-readable marking of AI‑generated content - Use imperceptible watermarks interwoven within content - Adopt a digitally signed “manifest/provenance certificate” for content that can’t securely carry metadata - Offer free detection interfaces/tools, including confidence scoring, and complementary forensic detection that does not rely on active marking - Test against common transformations and adversarial attacks - Use open standards and shared/aggregated verifiers to enable cross-platform detection and lower compliance friction The second section covers labelling deepfakes and certain AI-generated or manipulated text on matters of public interest and is applicable to deployers of generative AI systems, including: - Deepfake labelling - Modality‑specific labelling rules for real-time video, non-real-time video, images, multimodal content, and audio-only - Operational governance: encourages internal compliance documentation, staff training, accessibility measures, and mechanisms to flag and fix missing/incorrect labels.

We're observing a material shift in referral traffic sources, specifically an increase attributable to AI-driven entities. Understanding the mechanics of this traffic is crucial for effective site management and optimization. Server log analysis surface User-Agent string revealing visits from a new cohort of crawlers associated with large language model (LLM) based applications. These aren't your traditional search engine spiders exclusively focused on indexation. We're talking about agents representing services like Google's own AI features, Anthropic's Claude, Microsoft's Copilot integrated with Bing, and Perplexity AI, among others. These bots are actively fetching content, driven by user queries within their respective AI interfaces, which subsequently results in a referral back to the source URI. Quantifying this requires granular analysis of your access logs. By filtering and aggregating requests based on identifying User-Agent patterns, you can establish a baseline metric for AI bot visit frequency and volume. This data is invaluable for understanding the impact of these agents on your infrastructure and content reach. Controlling how these agents interact with your site is where the robots.txt protocol becomes paramount. It's about preventing indexation anymore and/or managing resource allocation and guiding specific user agents to appropriate sections of your site. Implementing well-defined Disallow and Allow directives, potentially even leveraging User-Agent specific rules, allows you to curate the crawl behavior of these diverse AI entities. This ensures that helpful agents, those that contribute to content visibility and referral traffic, can operate efficiently, while simultaneously mitigating potential issues with overly aggressive or unwanted scraping that could strain server resources or expose sensitive data. In essence, the rise of AI search bots necessitates a more sophisticated approach to log analysis and crawler management. It's a technical challenge that requires understanding the nuances of how these new agents identify themselves and interact with web resources to properly harness their potential for traffic generation while maintaining site integrity. Here's a great list of AI crawlers: https://lnkd.in/gZSgr4QC

🇪🇺 Everyone in AI governance is talking about Article 50 of the EU AI Act right now, and they should be. Article 50's transparency obligations became applicable on August 2: chatbots that disclose, deepfakes that get labeled, synthetic content marked in machine-readable form. But if your compliance radar only points at Brussels, you missed what happened in Sacramento on the exact same day. 🌴 California's AI Transparency Act (SB 942) went into force today too, as amended by AB 853. Here's what you need to know. 👇 🎯 Who's covered: a "covered provider" is a generative AI system producer with more than 1 million monthly visitors or users that is publicly accessible in California. That is not just the frontier labs. The duties: 🔍 A free public AI detection tool, with an API, so anyone can check whether content came from your system. 🏷️ Two kinds of content disclosure. Here's the keyword anchor I teach: latent is for machines, manifest is for humans. Latent disclosures are embedded in the content itself, machine-readable. Manifest disclosures are the visible label a user can choose to add. ⏱️ License hygiene: if you discover a licensee has stripped your latent-disclosure capability, you have 96 hours to revoke the license. ⚖️ Enforcement: $5,000 per violation, per day. And not just the Attorney General. City attorneys and county counsel can bring these actions too. 📅 What you can park for later: AB 853's own duty sets are not live yet. Large online platforms (over 2 million users) and GenAI hosting platforms come online January 1, 2027. Capture-device manufacturers follow January 1, 2028. Calendar them. No need to panic over them today. In practice, this means transparency compliance just became a two-continent problem with one shared deadline. If you're mapping Article 50 this week, map SB 942 right next to it. Same theme, same day, different enforcer. 🌍 Happy to be a resource if you're working through what this means for your organization. 🤝

Pandora’s Corpus: The Model Collapse Already Happening Inside Your Enterprise In 2023, Shumailov et al. (“The Curse of Recursion”) showed that when models train on model-generated data, the distribution collapses. Rare modes disappear first. The tails go silent. Now apply that inside your company. Your GenAI is generating executive summaries, CRM notes, underwriting commentary, code documentation, data annotations. At scale. Those artifacts land in SharePoint, Confluence, data lakes, vector stores. Within 12 to 24 months, synthetic and human-authored knowledge become indistinguishable. Then you fine-tune your next model on that corpus. That is recursive synthetic training inside your own enterprise boundary. What disappears first isn’t fluency. It’s edge-case judgment: the atypical claim, the nuanced regulatory interpretation, the exception underwriting logic. In regulated industries, that tail knowledge is where capital, compliance, and litigation risk live. The model won’t signal degradation. Your dashboards won’t turn red. Collapse is distributional, not dramatic. Three controls. Architectural, not aspirational. 1. Immutable provenance at write-time. AI-assisted artifacts must carry enforced, machine-readable origin metadata at the storage layer. 2. Synthetic exclusion partitions. Physically separate primary human corpora from AI-generated content. Training pipelines must default to exclusion. 3. Protected tail domains. Ring-fence human-validated underwriting, adjudication, and regulatory logic from recursive ingestion. Speed was correct. Uncontrolled recursion is not. #ModelCollapse #AIGovernance #EnterpriseAI

🗞️ 🇪🇺 Last reads for 2025! Great one on the future of gen-AI & traceability: Draft Code of Practice to make AI content technically traceable & perceptible to humans, while protecting trust, democracy & the information ecosystem! 👉🏼The Code proposes a full-stack transparency regime: • Technical traceability by default (providers) • Human-facing disclosure at the point of consumption (deployers) • Harmonised EU signals (taxonomy + icon) • With democracy & info integrity as explicit policy goals 🖊️ Written to operationalise Article 50 of the #AI Act providing a concrete framework to safeguard democratic trust & integrity of the information space in the age of #genAI. 🧭The doc focuses on AI-generated or AI-manipulated content (text, image, audio, video) for : 🔹Providers (technical marking & detection) 🔹Deployers (visible disclosure) 🛠️ Key proposal for providers: mandatory technical marking 🔹make all AI-generated or manipulated content machine-readable, detectable, robust, reliable, interoperable 🔹Require multi-layered marking required (single technique not enough): • Metadata with cryptographic signatures • Imperceptible watermarks embedded in content • Fingerprinting or logging where needed 🧩 Model-level responsibilities Make sure Generative AI model providers: • Embed marking at model level (esp. foundation models) • Support downstream deployers’ compliance • prohibit removal or tampering of marks in terms of use. 🧬 From labels to provenance chains Strong push for content provenance chains, not just AI/non-AI labels: • Record each AI and human modification step • Synchronise markings across text, image, audio, video • Providers encouraged to also support provenance for human-authored content. 🔍 Detection obligations for providers • offer free detection tools (API or public interface): • Confidence scores • Human-understandable explanations • Accessibility compliance • Long-term goal: shared / aggregated EU verifiers. 👁️ Obligations for deployers & #EU harmonisation 🔹clearly disclose AI-generated or manipulated content: Deepfakes (image, audio, video), AI-generated or manipulated text on matters of public interest. Disclose at first exposure. 🇪🇺 Common EU taxonomy & icon: Introduction of a shared taxonomy 🔹Fully AI-generated, showing degree of AI involvement & explaining what exactly was generated or manipulated; including audio-only & accessibility features Context-sensitive disclosure rules 🔹Detailed rules per format 🔹non-intrusive Disclosure for Creative, artistic, satirical works 🔹Editorial exception for text: No disclosure if human review + editorial responsibility are documented. 📄 Document Prepared by the @European Commission, led by the AI Office & DG CONNECT, with contributions from independent experts, industry, civil society, and academic stakeholders. 🗞️ enjoy the read ! 👏🏼 Kalina Bontcheva Dino Pedreschi Christian Ries Anja Bechmann Giovanni De Gregorio Madalina Botan

AI is reviving fundamental questions about how information is valued. In the late ’90s, we saw what happened when the ethos that “information wants to be free” collided with journalism, music and film. Business models collapsed. It took decades to rebuild a sustainable foundation for creators and publishers. GenAI is now testing that foundation again. Too much content is being scraped and used to train models without a reasonable exchange of value going back to the people and companies that created it. If we don’t fix that, we don’t just put individual businesses at risk – we weaken the information ecosystem that powers creativity, commerce and democracy. Together with Almar Latour and Dow Jones I wrote for the World Economic Forum about how we protect our information ecosystem in the AI era. Our view is simple: thoughtful regulation matters, but we can’t wait for it. At Pinterest, that means defending creators’ rights, giving people more control over how AI is used, and investing in safety and governance from the start. If we want AI to expand opportunity instead of eroding trust, we have to build those safeguards now. You can read our full piece here: https://lnkd.in/gSKuX9Yq

AI regulation isn’t settling, it’s reacting. And the reaction? Fragmented, global and and driven by public tension. Europe: The landmark AI Act is already under review. Why? Industry pushback. Now, the EU is signalling it may ease compliance and reduce red tape. United States: The proposed “AI Diffusion Rule” was pulled just before rollout. The focus has shifted from enforcement to diplomacy. China: Governance is tightening. The details remain unclear, but the intent is unmistakable: more control. It might seem like regulation is shaped only by politics, policy, and industry pressure. But now add the ethical and public concern layer. You don’t need expert analysis. Just read the headlines: →The New York Times is suing OpenAI over training data and copyright boundaries. →A GDPR complaint accuses ChatGPT of generating false, defamatory information. →A U.S. federal judge ordered OpenAI to preserve all ChatGPT outputs, marking a legal shift in how AI content is treated. Three regions. Three agendas. But one emerging pattern: → Public tension surfaces first, whether political, economic, or ethical. → Legal systems scramble to respond. → Governance becomes the tool to contain the risk. So what does this mean for leaders building with AI? If your strategy skips ethical alignment, regulation will catch you off guard. Ethics builds trust. And to navigate today’s grey areas and stay ready for shifting governance, you need to build with adaptability, documentation, and decision traceability in mind. Ethics is the why. Governance is the how. And both are becoming non-negotiable. 👇 How are you preparing for this dual front, ethical accountability and regulatory complexity? Sources in comments