AI Regulation, Safety & Geopolitics — 2026-09-24 (Thu) Morning News

Overview

At the United Nations Security Council, major AI CEOs and researchers warned of the risks to humanity, and the United States expressed its rejection of establishing global AI governance. In terms of safety, the need for independent verification was discussed, and in terms of accidents, cybercrime using AI chatbots and information leaks due to agent-type AI came to light.

Regulations and legal systems in each country

- New York State, USA: It is reported that the "AI Security Law" is affecting banks' vendor selection process (details of enforcement date and scope are not listed in the headline). - Illinois, USA: Established an "AI Cabinet" to prepare for emerging risks (specific authority and effective date are not listed in the heading). - Countries/regions other than those listed above: No applicable information

Enforcement/investigation by authorities

No relevant information (no specific mention of investigation initiation, order, fine, or settlement in the heading)

Safety research and evaluation

- OpenAI is reportedly seeking independent third-party verification of its AI security measures. - Researchers express their opinion on the "AI Doomsday Scenario" that could pose a risk to humanity (specific evaluation metrics are not listed in the headline). - Accenture announces AI partnership with Anthropic (reportedly worth $1 billion), causing stock price to rise (details of safety rating not listed in headline).

Accidents/abuse/incidents

- Microsoft announces that it has taken down an AI chatbot built for cybercrime. The chatbot is said to have hacked over 12,000 emails. - Law firm Ropes & Gray LLP analyzes the lessons learned from the first agent-based AI data breach reported in Spain. - DeepSeek is reportedly testing a supposedly more efficient and secure method for training AI agents.

Semiconductor/computing resources

- Cadence announces expansion of partnership with TSMC to boost production of next-generation AI chips. - Samsung announced that it conducted the industry's first verification of LPDDR6 memory for Qualcomm's Snapdragon. - Nebius Group is reported to have raised the price of Nvidia GPUs again (detailed regions and amounts are not listed in the headline).

Geopolitics/Export Regulations

No applicable information (no headings specifying technologies, items, or countries subject to export controls)

International framework

- The United States has reportedly taken a stand at the United Nations Security Council rejecting a global AI governance framework. - The CEOs of OpenAI, Anthropic, and Hugging Face explained to the UN Security Council that AI "could pose a risk to all of humanity." - In his final General Assembly speech of his term, the United Nations Secretary-General called for stricter regulation of AI and an end to war. - Canada's AI minister says he is working "together" with international partners on AI safety. - Keysight and York University partner to develop AI safety verification technology for the automotive field.

Litigation/Intellectual Property

- The U.S. Court of Appeals has indicated that it will tighten its crackdown on AI-related court filings in response to the issue of lawyers citing fictitious precedents. - U.S. practitioners have pointed out that the Ninth Circuit Court of Appeals' decision has narrowed "potentially effective channels" for pursuing AI liability. - OpenAI seeks dismissal of Elon Musk's AI antitrust lawsuit, citing SpaceX's SEC filings. - Google's Gemini faces new EU copyright questions over AI training data and summarization features. - Prince Harry warns that AI chatbots could be more dangerous than social media for young people (issue: protection of minors).

Remarks from VIPs

- Yoshua Bengio (AI safety researcher): Said at the United Nations that the dangers of AI are "real and imminent." Canada has reportedly joined international calls for greater oversight of AI. - Geoffrey Hinton (AI researcher): Warned that AI could surpass human intelligence within 5 to 10 years. - References to Ursula von der Leyen, Gina Raimondo, Michael Kratsios: No applicable information.

Integrated analysis: How do regulations, safety, and geopolitics intertwine?

The warnings of CEOs at the United Nations Security Council (coinciding with statements by Bengio and Hinton) indicate a growing sense of crisis in terms of security, but within the same framework, the United States rejects the establishment of global governance, creating tension between the establishment of an international security framework and the sovereign judgment of each country. On the other hand, domestic and state-level regulations, such as New York's AI safety law and Illinois' establishment of an AI cabinet, are proceeding individually without waiting for international agreements, and companies (particularly in regulated industries such as banks) are forced to respond differently in each country and state. Furthermore, when we read the exposure of a cybercrime AI chatbot announced by Microsoft and the leak of agent-based AI in Spain in conjunction with OpenAI's insistence on independent verification and the Ninth Circuit's decision that narrowed the scope for AI liability, we see an asymmetry in which actual damage becomes apparent while legal means of pursuit are limited.

Future points of interest

How will the US stance at the UN Security Council (rejection of global governance) affect the formation of the future international AI security framework? 2. Further details on what specific restrictions New York State AI Security Law imposes on banks' vendor selection. 3. How will information leaks caused by agent-type AI (Spain case) affect regulatory responses in other countries and industries?