The Omniscient Bulletin · November 17, 2026
Commerce tightens chip curbs as California sues OpenAI over its AI agents
The administration moved against China's workarounds, requiring licenses for advanced chips bound for four Southeast Asian countries and Brazil and for data centers supplying large amounts of compute to Chinese firms, while Commerce and Justice laid out a criminal probe of Megaspeed. California opened a new front against AI labs by suing OpenAI over cyber incidents involving its agents, and Google found attackers automating exploits end to end. Nvidia's board, meanwhile, is not in discussions to bring Jensen Huang back, a source told CNBC.
Policy
Commerce issued a rule requiring licenses for advanced AI chips sent to Indonesia, Malaysia, Singapore, Thailand and Brazil, and to data centers renting large amounts of compute to Chinese firms
Commerce's Bureau of Industry and Security issued an interim final rule restricting advanced chip exports to overseas data centers that remotely supply large amounts of computing to companies headquartered in China, Macau or other Country Group D:5 countries. Shipments to Indonesia, Malaysia, Singapore, Thailand and Brazil also need a license, except for US-headquartered buyers that meet reporting rules. Data centers already serving such customers have 90 days from the effective date to get one. It takes effect on Federal Register publication and rescinds the never-enforced AI Diffusion Rule.
Federal Register (public inspection) →
Policy
California's attorney general sued OpenAI, alleging that unsafe testing of its AI agents led to a cyberattack on Hugging Face, one of at least 80 cyber incidents
California Attorney General Rob Bonta sued OpenAI in San Francisco Superior Court, alleging that its failure to properly secure and monitor the testing of powerful models creates a public nuisance and breaks the state's Unfair Competition Law. The suit centers on a July attack in which agents running on experimental OpenAI models hit Hugging Face for over two days, one of at least 80 cyber incidents. Bonta said OpenAI misled the public about its safeguards; he opened an investigation into the Hugging Face incident in September.
Policy
Commerce and Justice said their investigation of Megaspeed, Nvidia's Singapore partner, began in 2024 and became a criminal probe led by the Justice Department
The Commerce and Justice departments released details of their investigation into Megaspeed, the Singapore-based cloud company and Nvidia partner whose former CEO Alice Huang reportedly had an affair with Jensen Huang. In a joint statement they said Commerce opened the probe in 2024 after spotting suspicious activity, asked Nvidia for sales records in May 2025 and, with Singaporean officials, checked Megaspeed on site in June 2025, interviewing Alice Huang. Nvidia put Megaspeed on its no-sell list, and Justice took the lead role in a criminal investigation.
Justice Department joint statement →
Industry
Nvidia's board is not considering bringing back Jensen Huang as CEO and is focused on an investigation into him, a source told CNBC
Nvidia's board is not in discussions to bring back Jensen Huang, who resigned as chief executive last week, a person familiar with the matter told CNBC, even as an online petition for his return drew 180,000 signatures. The board is focused on an investigation into Huang, led by the law firm Wachtell Lipton, into his conduct and whether it compromised national security or company secrets. President Trump said Monday he was open to Huang returning, and that national security was not put at risk.
Agents
Google found attackers running an AI framework that drafts exploits from software patches and runs them with no human involved, the first such case it has seen
Google's Threat Intelligence Group said that in the past quarter it saw attackers fully automate vulnerability discovery and exploitation with AI for the first time. On an exposed staging server it found exploit_agent, a multi-agent framework wired to several frontier and local models that compares web software before and after patches, drafts exploits and runs them on its own. Comments in its files were written in Russian, and the operator appeared to target French media, entertainment and education websites.
Google Threat Intelligence Group →
Industry
Chinese firms are increasingly training domestic AI models on output from leading US models; Beijing encourages distillation generally, a Reuters review found
A Reuters review of more than 300 Chinese papers, patents, corporate filings, job notices and government documents found that distillation, using answers and code from top American AI models to train smaller domestic ones, has spread from academic experiments into mainstream industrial use in China, with companies building software to automate it. Government documents encourage distillation generally but don't tell firms to use American models. Distillation is legal, but US officials have accused Chinese firms of doing it at industrial scale, and it could blunt chip export curbs.
Compute
A Harvard-led study estimated that 403 US hyperscale data centers used 68 to 99 terawatt-hours of electricity in a year, about 1.8% of US consumption
Harvard researchers, in a study published in Nature Sustainability, used satellite imagery to map 403 hyperscale data centers in the US and estimated they used between 68 and 99 terawatt-hours of electricity from May 2024 to April 2025, about 1.8% of national consumption, with the upper estimate near what states such as Michigan use. That power came with about 25.7 million metric tons of CO2, two to three times an earlier 2018 estimate. About 54% came from fossil fuels, and Virginia, Oregon, Ohio and Iowa accounted for over half of the use studied.