Private

Outcome/discussion topic 1: AI Terminology - "a cat is a cat" Artificial Intelligence borrows heavily from medical and psychological lexicons (e.g., "intelligence," "hallucination"). This humanization of machines creates confusion and fear, as the terminology is misleading and inappropriate. In medical fields but not limited to, it hinders adoption, fuels conspiracy theories, and erodes trust. Proposed solution: Enforce that all organizations and corporations use standardized, neutral terminology for AI and its behaviors. An official, universally adopted lexicon would break the current misleading pattern, bring clarity, and potentially slow down aggressive tech marketing strategies. Outcome/discussion topic 2: AI Data Training & Ethical Agreements There is a critical lack of transparency regarding what data is used to train Large Language Models (LLMs), including potential copyright infringement and unauthorized use of personal data. Organizations have been pressured to migrate their data to the cloud ("a great business model: pay to have your data potentially stolen and used for LLM training"), exposing everything to risk. Proposed discussion point: Establish binding ethical agreements and data provenance standards for AI training, including opt-out mechanisms, audit trails, and legal accountability for unauthorized data use. Outcome/discussion topic 3: Intellectual Property and AI. Patent offices worldwide are becoming misaligned in their handling of computer related inventions including AI-related inventions. Large tech organizations have been influencing patent office approvals and boards of review. If left unchecked, this could have long-term dramatic consequences for the economy and public trust in the IP system. Proposed discussion point: Harmonize international patent examination standards, ensure independence of review boards from industry influence, and reassess patent eligibility criteria for AI, AI-generated or AI-assisted inventions. Outcome/discussion topic 4: Sponsoring for Creation of the Dataterms Analogous to Incoterms in international trade, the time has come to develop a robust, standardized system for data trading. This framework should be established by governmental institutions (not solely private corporations) to create the foundation for a trustworthy, transparent data economy of the future. Proposed discussion points: Define Dataterms as binding legal and technical standards for data ownership, transfer, licensing, and usage rights. Clarify obligations and liabilities when data is sold, shared, or used for AI training. Include provisions for consent, revocation, audit trails, and cross-border data flows. Prevent lock-in by big tech by ensuring interoperability and public oversight. Alignment with prior outcomes: Outcome 1 (terminology): Dataterms would provide precise language around data transactions, reducing misleading marketing claims. Outcome 2 (training data ethics): Dataterms would mandate transparency about whether data can be used for LLM training. Outcome 3 (IP & patents): Dataterms would clarify who owns data-derived IP, especially when AI generates outputs from traded data.

Social, economic, ethical, cultural, linguistic and technical implications of AI;Interoperability of governance approaches;Safe, secure and trustworthy AI;AI capacity-building;

Alignment with prior input (question1): Outcome 1 (terminology): Dataterms would provide precise language around data transactions, reducing misleading marketing claims. Outcome 2 (training data ethics): Dataterms would mandate transparency about whether data can be used for LLM training and generate revenues to their Data Owners Outcome 3 (IP & patents): Dataterms would clarify who owns data-derived IP, especially when AI generates outputs from traded data.

I have the feeling I already explained it above. Thank you!

The speed of implementation, lack of interoperability (including on intellectual property) and dependency to private organization are bringing everyday bigger social gaps in population on top of fundamental knowledge losses. This affects all of us on short, mid and long terms.

The AI Dialogue's real role is to move from abstract principles to enforceable international standards — on terminology, training data transparency, patent alignment, and a new 'Dataterms' framework for data trading. That's the work I'm ready to do.

Existing initiatives the AI Dialogue should build upon: OECD AI Principles: Trustworthy AI by design, but lacks enforcement. UNESCO Recommendation on AI Ethics: Strong on human rights, weak on technical standards. GPAI (Global Partnership on AI): Good for multi-stakeholder dialogue, but produces mostly non-binding reports. EU AI Act: First binding regulatory framework, but regional, not global. WTO e-commerce negotiations: Includes data governance, but avoids AI training data specifically. ICANN / WIPO: Models for multi-jurisdictional coordination (naming/patents), but nothing analogous for data trading. Added value the AI Dialogue could bring: Unlike existing bodies: which are either too vague (OECD, UNESCO), too regional (EU AI Act), or too industry-friendly (many multi-stakeholder forums). The AI Dialogue would: 1.Bridge terminology chaos (Q1.Outcome 1) by adopting an official, medically-grounded AI lexicon. 2.Close the LLM training data loophole (Q1.Outcome 2) with binding transparency and consent protocols. 3.Realign patent offices globally (Q1.Outcome 3) through a coordinated review standard. Create "Dataterms" (Q1.Outcome 4) – an Incoterms-style framework for data trading, governed by states, not corporations. In short: existing initiatives set principles. The AI Dialogue can deliver operational standards, turning goodwill into governance.

Take accountability and engage in solid actions after the Dialogue session.

Note: I would be delighted to bring and explain my vision further.

Innovative engagement formats for the AI Dialogue: 1. "Red Team / Blue Team" Regulatory Sandboxes Live, adversarial role-play. One team (Big Tech) tries to exploit terminology loopholes (Outcome 1) or data training gaps (Outcome 2). The other team (regulators / civil society) must close them in real time. Audience votes on solutions. Produces battle-tested policy drafts. 2. Patent Office Reverse Simulation Invite international patent litigators, examiners, judges, and tech lobbyists to argue opposite sides of an AI patent case (Outcome 3). Forces transparency around influence tactics. Output: a public "influence map" and draft harmonized examination guidelines. 3. Dataterms Negotiation Bootcamp Modeled on UN trade law negotiations (UNCITRAL style). Assign countries, tech firms, and data cooperatives. Negotiate one clause of "Dataterms" (Outcome 4) per session — e.g., "Who owns data used for LLM training?" Live text editing on screen or with the help of Miroki as note taker for instance. Output: a living, contested, but converging draft. 4. "Missing Glossary" or "AI replacement terminology" Workshop Crowdsource the worst misleading AI terms (e.g., "intelligence", "hallucination"). Then, in a structured dialogue, replace each with precise, neutral alternatives (Outcome 1). Output: a public, adoptable lexicon with dissenting opinions noted — not just consensus-washing. 5. Whistleblower & Data Donor Panels (closed then public) Anonymized testimony from data workers, content moderators, and individuals whose personal data was used for training (Outcome 2). First closed for safety, then summarized publicly. Adds moral weight to technical debates. Added value: These formats replace passive listening with active friction — exposing power asymmetries and producing actionable outputs, not just declarations.

See previous answers. Thank you