AI Regulation, Safety & Geopolitics — 2026-11-07 (Sat)

Overview

Regulation: A court adviser in the EU said AI chatbots that replace reading the news may breach publisher rights, and UK ministers are seeking powers to investigate AI agents (Dealroom, Cape Argus, The Register). Safety: OpenAI launched GPT-6 Cyber, and METR found AI is likely doubling Anthropic's R&D pace (The Hacker News, Resultsense). Geopolitics: Moonshot is seeking Nvidia chip access in Europe, and Von der Leyen said Canada will train AI on EU supercomputers (CNBC Television, EU Perspectives). 【National and regional regulation】 - EU: A court adviser said AI chatbots that replace reading the news may breach publisher rights; this is an adviser's view, not a ruling (Dealroom). Another report says a Google AI chatbot may break press copyright rules, according to an EU legal adviser (Cape Argus). - UK: Ministers are seeking powers to investigate when AI agents are let off the leash; this is a request for powers, not an enacted law (The Register). - US: The OCC issued an examiner framework for banks' use of artificial intelligence (JD Supra).

Enforcement and investigations

- The OCC issued an examiner framework for banks' use of artificial intelligence (JD Supra). - UK ministers are seeking powers to investigate AI agents (The Register). - Three people were charged in an AI scheme used to impersonate CEOs and attorneys to steal $129 million (New York Post).

Safety research and evaluation

- OpenAI launched GPT-6 Cyber after AI helped find 81 Firefox flaws and 948 Linux exploits (The Hacker News). - METR found AI is likely doubling Anthropic's R&D pace (Resultsense). - A study compared medical AI benchmarks with real clinical workflows (Telehealth.org). - Asia Society published a piece arguing India should build an interoperable AI safety mechanism in BRICS (Asia Society). - Anthropic's IPO faces a valuation cut after a $2.5 trillion rout (The Jerusalem Post).

Incidents and misuse

- SC Media discussed what the OpenAI–Hugging Face incident says about AI agent access (SC Media). - Three people were charged over an AI scheme that impersonated CEOs and attorneys to steal $129 million (New York Post). - Living Security said preemptive security starts with visibility into human and AI agent risk (Newswire).

Semiconductors and compute

- TSMC is mulling making advanced AI memory chips, according to Information (Bloomberg). - Nvidia tapped TSMC to build 3D DRAM, which threatens to reshuffle the memory supply chain (BigGo Finance). - Nvidia is eyeing Apollo's struggling telecom sites to sidestep data center moratoriums, according to a report (Yahoo Finance, The Information). - A commentary argues the balance of power between AMD and Nvidia is shifting to AMD (24/7 Wall St.). 【Geopolitics and export controls】 - Chinese AI lab Moonshot is seeking Nvidia chip access in Europe (CNBC Television). - Von der Leyen said Canada will train AI on EU supercomputers (EU Perspectives). - Asia Society argues India should build an interoperable AI safety mechanism in BRICS (Asia Society).

International frameworks

- Takeaways from IAPP P.S.R. + AI Governance Global 2026 point to a desire for AI operational standards, a shifting lens for analyzing third-party AI services, and continued focus on state privacy laws (Ashurst Perkins Coie). 【Litigation and intellectual property】 - Parents sued Google over a teen's AI-assisted suicide planning; the case concerns the suicide of 13-year-old Emma (Punch Newspapers, Organización Mundial Bullying Sin Fronteras). - The EU court adviser's view on AI chatbots and publisher rights concerns press copyright (Dealroom, Cape Argus). - The LCJ warned that AI use by litigants in person is adding to the courts' burden (The Law Society Gazette).

Statements by key figures

- Ursula von der Leyen (European Commission President): said Canada will train AI on EU supercomputers (EU Perspectives). - Geoffrey Hinton (AI researcher): said humans may regret creating AI and that humanity might have been better off without it (Cybernews, News18). He also warned AI could cause a major blackout by 2027 (Mashable India). 【Integrated analysis: what is happening】 Regulation is moving toward agents and content. The UK's request for investigation powers over AI agents and the OCC's examiner framework for banks target how AI is used, while the EU adviser's opinions target how chatbots use press content (The Register, JD Supra, Dealroom). These appear to be converging with safety signals: the OpenAI–Hugging Face incident raises questions about agent access, and GPT-6 Cyber follows AI-assisted discovery of Firefox and Linux flaws (SC Media, The Hacker News). Companies deploying agents may need to document access controls and be ready for supervisory questions.

On compute, Nvidia's search for sites that sidestep data center moratoriums suggests local opposition is shaping where capacity gets built (Yahoo Finance, The Information). Moonshot's search for Nvidia chips in Europe and Canada's planned use of EU supercomputers indicate that access to compute is becoming a cross-border arrangement (CNBC Television, EU Perspectives). This is inference from the headlines.

For users and buyers, litigation over chatbot harms and publisher rights, together with the fraud case, may raise the compliance and liability weight of consumer-facing AI products (Punch Newspapers, New York Post, Cape Argus).

What to watch

Whether the EU court follows its adviser's view on chatbots and publisher rights, which may shape how AI services use news content. 2. Whether UK ministers obtain the investigation powers over AI agents they are seeking. 3. Whether Nvidia's reported plan for Apollo's telecom sites and its 3D DRAM work with TSMC proceed, given the data center moratoriums.