AI Policy · Daily

President Trump told APEC leaders in Shenzhen that Washington will reject international limits on AI, saying the U.S. opposes "any form of globalist regulation" of superintelligence while acknowledging the technology has "some bad elements." Anthropic begins trading on the Nasdaq today under the ticker ANT, after pricing its initial public offering at $83 a share, above the planned $70 to $78 range, raising roughly $81 billion and valuing the Claude maker at about $1.5 trillion. Chinese President Xi Jinping used a meeting with Trump on the APEC sidelines to press for coordination on AI safety risks and to say the technology should "always remain under human control," calling the talks "positive and constructive." The U.S.-China Economic and Security Review Commission told Congress in its annual report that the People's Liberation Army is seeking Nvidia's most advanced processors for AI-enabled weaponry and that Beijing is working to evade U.S. export controls.

I.Top Stories

Trump vows at APEC to fight international AI regulation

President Trump told leaders at the Asia-Pacific Economic Cooperation summit in Shenzhen on Thursday that the U.S. will not accept international limits on AI. "We oppose any form of globalist regulation or anything to control and stifle SI, and we won't allow that to happen, frankly," Trump said, using his shorthand for superintelligence. He also said AI has "some bad elements" that must be managed carefully.

Read at Bloomberg ↗ • Read at CBS News ↗

Anthropic begins trading on Nasdaq today after pricing IPO at $83 a share

Anthropic begins trading on the Nasdaq today under the ticker ANT, after pricing its initial public offering at $83 a share, above its initial $70 to $78 range, per AP. The price values the maker of the Claude chatbot at about $1.5 trillion. The roughly $81 billion in proceeds tops the approximately $75 billion SpaceX raised in June. The offering covers 980 million shares, about 956 million of them sold by the company and the rest by existing stockholders. Anthropic has signed leases and other deals committing it to $900 billion in computing and infrastructure spending in coming years, including $70 billion over the next 12 months.

Read at AP ↗ • Read at CNBC ↗

Xi presses Trump to keep AI under human control as Beijing calls talks positive

Chinese President Xi Jinping called his meeting with President Trump on the APEC sidelines in Shenzhen "positive and constructive," per Bloomberg. Xi said China and the U.S. "should strengthen communication and coordination to guard against artificial intelligence safety risks and ensure that artificial intelligence always remains under human control," Hong Kong public broadcaster RTHK reported. Before the meeting, administration officials told Wired that an AI incident hotline announced in September, linking Treasury Secretary Scott Bessent with Chinese Vice Premier He Lifeng, remains at the concept stage and is not operational. Officials said U.S. worries center on autonomous models slipping human oversight, while Beijing is more concerned about dissidents using open-weight modelsOpen-weight modelAn AI model whose trained parameters are published so anyone can download, run and modify it, as opposed to one reachable only through the developer's API. The main policy tension is that open weights spread capability widely and cannot be recalled once released., which anyone can download and modify.

Read at Bloomberg ↗ • Read at RTHK ↗

Congressional commission finds China's military pursuing top Nvidia chips around U.S. export curbs

The U.S.-China Economic and Security Review Commission, a bipartisan panel that reports to Congress, said in its annual report that the People's Liberation Army wants Nvidia's most advanced processors for AI enabled weaponry, per Reuters. Beijing is trying to evade U.S. export controls to fuel its drive to build AI into weapons, command systems and logistics, the commission found. Chips designed by China's Huawei and Cambricon still trail those from Nvidia and Advanced Micro Devices, according to the report. Chinese companies have also used Nvidia chips remotely through data centers in third countries such as Malaysia. The Commerce Department issued a rule Monday meant to stop that kind of remote access from getting around existing export limits.

Read at Reuters ↗ • Read at Quartz ↗ • Read at USCC ↗

DHS Secretary Mullin, DNI Clayton tell House panel outages were maintenance problems, not attacks

Homeland Security Secretary Markwayne Mullin, Director of National Intelligence Jay Clayton and FBI Deputy Director Christopher Raia told the House Committee on Homeland Security that the recent outages at airlines, banks and government agencies were neither connected nor malicious, The Washington Post reported. Clayton called them "separate service outages associated with maintenance and technical changes" and said they were "not a coordinated or state-sponsored attack." Lawmakers from both parties questioned the officials at the hearing, titled "Worldwide Threats to the Homeland: 25 Years After 9/11," per The Hill. Bank executives and NSA officials had privately told lawmakers the outages came from rushed fixes to security holes that AI models found, as reported by The Washington Sun in AIPD's November 18th edition.

Read at The Washington Post ↗ • Read at The Hill ↗ • Read at House Homeland Security Committee ↗

Huang agrees to testify publicly before Senate Banking

Former Nvidia CEO Jensen Huang has agreed to appear in public before the Senate Committee on Banking, Housing, and Urban Affairs, a source in Chairman Tim Scott's office told Yahoo Finance. Commerce Secretary Howard Lutnick, as previously reported, has also agreed to testify. Scott (R-S.C.) said on CNBC that he wants the testimony finished before Christmas, and talks over timing are continuing. Senators are expected to ask whether the administration did enough to stop apparent attempts to evade U.S. export controls on advanced chips. Huang is also scheduled for a closed-door, transcribed interview Dec. 1 with the House Committee on Foreign Affairs, the House Permanent Select Committee on Intelligence and the House Select Committee on the Chinese Communist Party. Nvidia general counsel Tim Teter is to follow in a separate closed session the next day.

Read at Yahoo Finance ↗ • Read at CNBC ↗

II.China Watch

Alibaba backed Megaspeed and co-founded military linked AI lab, NYT finds

Alibaba was the main customer and financial backer of Megaspeed, the Singapore company that bought more than $4 billion in Nvidia chips, The New York Times reported, citing three people familiar with the relationship. Alibaba also co-founded Zhejiang Lab, an AI research center with close ties to China's military whose researchers have worked on automating cyberattacks and simulating war games. Whether chips supplied by Megaspeed moved from Alibaba to the lab is unclear. The lab's director, Wang Jian, is Alibaba's former chief technology officer and a longtime friend of Jensen Huang. An Alibaba spokeswoman rejected any suggestion that it serves the military and said its cloud service agreement bars military use of its products.

Read at The New York Times ↗

Jury begins weighing U.S. racketeering case against Huawei after two month trial

Jurors in federal court in Brooklyn began deliberating in the U.S. criminal case against Huawei Technologies, the South China Morning Post reported. The Shenzhen-based company, the world's largest maker of telecommunications equipment, faces 12 counts including racketeering conspiracy, theft of trade secrets, bank fraud and obstruction of justice. Prosecutors allege a 20-year scheme in which Huawei leaders pushed employees to take proprietary technology from competitors and misled banks about the company's business in Iran. In closing arguments, defense attorney Brian Heberlig said prosecutors "decided Huawei was a criminal enterprise" and built their case from that premise, according to The Associated Press. Huawei and three affiliates have pleaded not guilty, and deliberations were set to resume Thursday.

Read at SCMP ↗ • Read at AP News ↗

III.Policy Tracker

Energy and Commerce Chair Guthrie asks Nvidia's Kress, Commerce Secretary Lutnick for chip export briefings

Rep. Brett Guthrie (R-Ky.), chair of the House Committee on Energy & Commerce, sent letters seeking briefings from interim Nvidia CEO Colette Kress and Commerce Secretary Howard Lutnick, per The Hill. Guthrie asked Kress to explain Nvidia's internal export compliance controls before Dec. 2, including whether the company helped Singapore's Megaspeed get around U.S. restrictions on AI chips. He wrote that the reported events raise questions about Nvidia's compliance with federal export law and the strength of its security controls. His letter to Lutnick sought details on how Commerce applies export controls to semiconductors. Guthrie also asked Jeffrey Kessler, the under secretary of the Bureau of Industry and Security, Commerce's export control arm, to appear.

Read at The Hill ↗

Sens. Cantwell, Rounds press AI czar Clayton to fix flaws AI can expose

Sen. Maria Cantwell (D-Wash.), ranking member of the Senate Committee on Commerce, Science, and Transportation, and Sen. Mike Rounds (R-S.D.), who chairs a Senate cybersecurity subcommittee, wrote to Jay Clayton on Tuesday asking him to identify and fix weaknesses in U.S. infrastructure, The Guardian reported. Clayton serves as both director of national intelligence and White House AI czar. The senators wrote that AI lets attackers find and exploit software flaws faster than many organizations can react. Even when a fix exists, rolling it out across linked systems can take weeks or months and interrupt essential services, they wrote.

Read at The Guardian ↗ • Read at Senate Commerce Committee ↗

Sen. Warren urges SEC, Congress to act as Anthropic goes public

Sen. Elizabeth Warren (D-Mass.), ranking member of the Senate Committee on Banking, Housing, and Urban Affairs, said in a statement from her office that "Trump's regulators are asleep at the switch" as Anthropic prepared to list its shares. She said the SEC must protect investors and Congress must move quickly to put safeguards on advanced AI development. Her statement cited a letter that leaders of public pension plans sent Anthropic the same day. In a letter shared with Axios, officials from five Democratic-led states, including Maryland, Illinois and New Mexico, asked Anthropic to explain how it will manage the catastrophic risks it has flagged and to set an end date for its founders' voting control.

Read at Senate Banking Committee ↗ • Read at Axios ↗

Senate clears another cloture vote on permitting bill with data center power terms

Senators voted 65-31 to limit debate on the Bipartisan American Affordability and Jobs Act, one of a series of procedural votes moving the permitting overhaul toward final passage, The Hill reported. The bill aims to speed construction of energy and other infrastructure. It would shorten approval timelines for both clean energy and fossil fuel projects and require data centers to pay for the power they consume, according to the Hill. House Minority Leader Hakeem Jeffries, a New York Democrat expected to become speaker in January, has said the measure falls short in several areas.

Read at The Hill ↗

New York, Los Angeles, Chicago districts eye Search alternatives as Google ties it to Gemini AI

The country's largest school districts, in New York, Los Angeles and Chicago, are considering alternatives to Google Search after the company said schools that switch off its Gemini AI features will lose Search as well, per The New York Times. Starting next year, schools will no longer be able to disable Gemini features while keeping Search. Google's vice president of education, Chris Phillips, said the two products are closely intertwined. New York City Schools Chancellor Kamar H. Samuels said the city's schools will look at other search options unless Google changes course. Chicago Public Schools has told Google it will seek other search tools if it cannot get Search without AI, The Verge reported. Google said Gemini, Gemini Notebook and Search will be off by default next year until each school chooses which to enable.

Read at The New York Times ↗ • Read at The Verge ↗

IV.Capability & Research Watch

Hackers push credential stealing versions of Microsoft's AI agent toolkit to PyPI

Attackers published two malicious versions of Microsoft Agent Framework, the company's open source toolkit for building and coordinating AI agentsAI agentAn AI system that carries out multi-step tasks on its own, such as browsing, writing code or making purchases, rather than answering a single prompt. Agents raise new questions about liability, security and oversight because they act rather than just advise., to PyPI, the main public repository for Python software, on Tuesday, BleepingComputer reported. The package draws about 200,000 downloads a month. Versions 1.26.1 and 1.26.2 ran a credential stealer whenever Python started, even if the framework was never imported, targeting secrets for Azure, AWS, GitHub and other services. PyPI quarantined the first release, but the attacker published the second about 2.5 minutes later, showing the publishing access was still live. Security firms Socket, StepSecurity and Ox Security documented the attack, and developers are advised to rebuild from the clean 1.26.0 release.

Read at BleepingComputer ↗

V.Industry & Market Watch

Nadella says much critical infrastructure can't be patched as AI speeds up flaw discovery

Microsoft CEO Satya Nadella told The New York Times' DealBook Summit that many internet connected systems underpinning U.S. infrastructure carry software flaws that cannot be fixed, per TechCrunch. "Remember, there is a lot of stuff in our critical infrastructure that's unpatchable," he said, citing firmware, data center cooling equipment and HVAC systems. Nadella said AI now gives organizations a powerful way to find vulnerabilities. Fixes must reach production almost as soon as they are released because attackers exploit flaws quickly, he said. He said that pace will cause instability. He attributed Microsoft's recent Outlook outage to a latent bug and said the company would fix both architectural and operational problems behind it.

Read at TechCrunch ↗

Citi, PNC, Citizens make customers liable for third-party AI agents, even when hacked

Citi updated its online user agreement so that customers who give third-party AI agents account access "are responsible for activities conducted through those agents," a spokesperson said. Under Citi's terms, customers are also responsible if an agent leaks their credentials or is hacked. PNC changed its terms the same week, and Citizens emailed customers warning that they are responsible for what their AI assistants do.

Read at TechCrunch ↗