AI Regulation, Safety & Geopolitics — 2026-11-04 (Wed)

Overview

Regulation: Turkey's competition authority opened an antitrust investigation into Google's AI search tools (PYMNTS.com). Safety: Barbados and the United Nations will host an AI safety summit, and Germany tied its AI safety institute to Canada (Digital Watch Observatory, Aju Press). Geopolitics: Nvidia reportedly told cloud clients to sell 51% of services outside China (Türkiye Today). 【National and regional regulation】 No notable developments.

Enforcement and investigations

- Turkey's competition authority (CPI) opened an antitrust investigation into Google's AI search tools (PYMNTS.com). The headline gives no further detail on scope or timing.

Safety research and evaluation

- No evaluation results, system cards or third-party verification appear in today's headlines. - Business Matters runs a piece headlined "AI Safety Has a Trust Problem. Chris Hsu Is Betting on Proof Instead." (Business Matters) - South Korean and Canadian experts discussed AI safety amid cyber threats (Aju Press). 【Incidents, misuse and security events】 - No specific incident with a named victim or responder appears in today's headlines. - Epstein Becker Green published "When the AI Agent Goes Off-Script: The First Incident Playbook for Frontier AI" (Epstein Becker Green). - A SEON survey found that 65% of U.S. consumers blame AI companies for agent-driven fraud (FF News). - Biometric Update argues that an AI agent's biometric approval needs an expiration clock (Biometric Update). - Vendors announced products: Chargebacks911 introduced AI Agent Purchase Protection for merchants (Supply & Demand Chain Executive), and PagerDuty debuted its Paige AI agent with a renamed PD Reliability Platform and PD Labs (SiliconANGLE).

Semiconductors and compute

- Fireworks contracted 5MW of Nvidia Blackwell Ultra capacity at the Era4 data center in the UK (Data Center Dynamics). - Samsung SDS broke ground on a 60-120MW AI data center in Gumi (Data Centre Magazine). Maeil Business reports that Samsung launched a $45bn physical AI drive with the Gumi AI data center (Maeil Business). - BigGo Finance's headline says Anthropic's S-1 commits $500-800 Billion to compute as Intel's capacity planning fails again (BigGo Finance). - The Information reports on how Nvidia uses its soft power to influence customers and shape policy (The Information). 【Geopolitics and export controls】 - Nvidia reportedly told cloud clients to sell 51% of services outside China (Türkiye Today). The headline does not say which services or products are covered, or whether any government required this.

International frameworks

- Barbados and the United Nations will host an AI safety summit (Digital Watch Observatory). - Germany tied its AI safety institute to Canada as Merz urged speed (Aju Press). - Naver secured ISO 42001 certification for its AI Tab and expanded its safety system in Korea (Chosunbiz). BigGo Finance describes the AI Tab as the first search-based agentic AI to earn ISO/IEC 42001 certification (BigGo Finance). - Johns Hopkins University lists a seminar on NIST and measurement for trustworthy tools in the age of AI (Johns Hopkins University). 【Litigation and intellectual property】 - A judge dismissed a civil suit filed by victims' parents in the Lancaster Country Day School A.I. scandal (FOX43). - A US appeals court ruled that a law firm's AI use does not warrant cutting legal fees (Reuters). The Times of India reports that a US company had argued lawyers using artificial intelligence should receive lower fees (The Times of India). - Elvis Presley enters the AI copyright battle 49 years after his death (Daily Sabah). - Hollywood's biggest players are backing the New York Times' copyright lawsuit against OpenAI (Page Six).

Statements by key figures

No notable developments. 【Integrated analysis: what is happening】 - Safety work is becoming institutional and cross-border. Germany's link with Canada, the UN-hosted summit in Barbados and Naver's ISO/IEC 42001 certification (Aju Press, Digital Watch Observatory, Chosunbiz) suggest that safety is moving from research toward standards and certification. Companies deploying agentic AI may increasingly face questions about certified management systems. - Agent-related risk is mostly framed through fraud and liability. The SEON survey, the Biometric Update piece and the Chargebacks911 product (FF News, Biometric Update, Supply & Demand Chain Executive) indicate that consumers place responsibility on AI companies. This may push merchants and AI providers toward incident planning and purchase-protection measures. - Compute buildout and geopolitics are linked. Data center commitments in the UK and Korea, alongside the reported Nvidia instruction on services outside China (Data Center Dynamics, Data Centre Magazine, Türkiye Today), appear to show capacity being placed in a way that reflects policy exposure. This is inference, as the headlines do not state the cause. - Competition enforcement is reaching AI products. Turkey's investigation into Google's AI search tools (PYMNTS.com) suggests that AI features inside existing products can come under antitrust scrutiny. Companies bundling AI into search may need to prepare for similar inquiries. - Courts are setting practical boundaries. The Reuters ruling on legal fees and the copyright cases (Reuters, Page Six, Daily Sabah) indicate that AI use is being tested in billing and in the use of creative works.

What to watch

- Whether Turkey's competition authority publishes the scope of its Google AI search investigation (PYMNTS.com). - Details of the Barbados and United Nations AI safety summit, and of the Germany-Canada safety institute link (Digital Watch Observatory, Aju Press). - Whether the reported Nvidia instruction on selling 51% of services outside China is confirmed or clarified (Türkiye Today).