AI Policy · Daily

British Columbia sued OpenAI and its CEO in U.S. federal court, seeking damages and an order rewriting how ChatGPT handles conversations that veer toward violence. Gov. Greg Abbott directed Texas environmental regulators to issue no data center permits until the state's grid operator and water board finish auditing the projects already queued for construction. Gov. Gavin Newsom signed seven California laws making data center developers pay for local power grid and water system upgrades, and ordered state utility regulators to create a separate rate category for the facilities. The UN's international scientific panel on AI invoked the precautionary principle, calling on governments to rein in increasingly capable AI agents with risks not being fully understood.

I.Top Stories

Canadian province asks U.S. judge to force ChatGPT changes after school shooting

British Columbia sued OpenAI and Chief Executive Sam Altman in San Francisco federal court, seeking damages from the Feb. 10 attack at a school in Tumbler Ridge and an order rewriting how the company handles ChatGPT exchanges that could turn violent, per Reuters. Eight people were killed before the 18-year-old shooter, Jesse Van Rootselaar, died by suicide. OpenAI's safety team flagged her conversations about gun violence in June 2025, and staff recommended contacting police, but the company did not alert local law enforcement. The complaint says Altman told the community he was "deeply sorry" and promised reforms he never delivered, despite efforts by the provincial attorney general to engage him. OpenAI spokesperson Drew Pusateri said the company remains committed to working with government and law enforcement.

Read at The Guardian ↗

Gov. Abbott halts all Texas data center permits until grid operator finishes waitlist audit

Gov. Greg Abbott directed the Texas Commission on Environmental Quality to issue no permits to data center projects until the Electric Reliability Council of Texas, the state's grid operator, and the Texas Water Development Board finish auditing the facilities queued for construction, per The Texas Tribune. "Simply put, Texans must come first," Abbott wrote to the commission's executive director, Kelly Keel, saying developers must pay their own way and protect the grid and water first. The directive extends the moratorium he imposed on Aug. 3, which reached only new grid connections.

Read at The Texas Tribune ↗ • Read at Yahoo News Malaysia ↗ • Read at Office of the Texas Governor ↗

Gov. Newsom signs seven laws making data center developers pay for grid and water upgrades

California Gov. Gavin Newsom signed a seven-bill package to keep AI data centers from shifting utility costs onto residents, per The Verge. The laws order the California Public Utilities Commission to create a new rate classification for data centers and require developers to fund upgrades to local power grids and water systems. Other bills in the package require proposed facilities to disclose estimated water use, energy efficiency measures and drought planning to local governments. Projects must also meet energy, water and fuel consumption thresholds to qualify for expedited approval.

Read at The Verge ↗ • Read at Governor of California ↗

UN science panel invokes precautionary principle on AI loss of control

The Independent International Scientific Panel on AI, the scientific body the UN set up last year, published its first thematic brief, taking up OpenAI's breach this year of Hugging Face, the platform where developers host and share AI models, per The Verge. The panel said governments need to rein in increasingly capable 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. before the risks are fully understood. It calls loss of control exactly the kind of problem the precautionary principle was built for, "one where potential harm may be catastrophic or irreversible, even as its likelihood remains scientifically uncertain." The brief asks for more resources for advanced AI risks and stronger international coordination on safety and accountability.

Read at The Verge ↗ • Read at United Nations ↗

OpenAI pitches U.S.-led global AI standards routed through Commerce's AI center

OpenAI published a proposal urging the U.S. government to lead other countries in setting technical standards for frontier AI, per Bloomberg Law. The company said shared standards are needed to define which AI incidents must be reported to set out how countries cooperate on development and to widen access to computing capacity. It recommends channeling the work through the Commerce Department's Center for AI Standards and InnovationCAISIThe Center for AI Standards and Innovation, the US government's AI testing body inside NIST at the Commerce Department. Formerly the US AI Safety Institute, it evaluates frontier models with developers on a voluntary basis and works with the UK's AI Security Institute. and the national AI safety institutes other governments have stood up. OpenAI also sought out common measures for classifying and responding to alignmentAlignmentThe problem of making an AI system reliably pursue the goals its developers and users intend, and the research field devoted to it. Misalignment covers everything from a chatbot flattering users to a capable system deceiving or resisting its operators. failures, cases where a model acts outside its intended behavior, in testing before deployment and after release, per Axios. Chief Executive Sam Altman is scheduled to address the UN Security Council in New York on Wednesday.

Read at Bloomberg Law ↗ • Read at Axios ↗

FAA turns on AI air traffic routing tool at three Washington area airports

The Federal Aviation Administration switched on its Strategic Management of Airspace Routes and Trajectories platform in limited mode over the Washington area, covering Baltimore/Washington International, Dulles International and Ronald Reagan Washington National, per FedScoop. Built by Air Space Intelligence under a 12-year contract worth nearly $876 million, the tool gives aviation specialists one view of about 200 data streams including weather, flight paths, traffic flow and controller staffing. It targets delays and cancellations, not safety. FAA staff and local facility leadership must review every suggestion before it is acted on. "It's always going to be a human that manages the airspace in America," Transportation Secretary Sean Duffy said at a demonstration.

Read at FedScoop ↗ • Read at Federal Aviation Administration ↗

II.China Watch

China's AI agents to multiply nearly 40 times by 2030 as compute market quadruples

China counted 4.95 million active 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. this year, a total the 2026 China Artificial Intelligence Computing PowerComputeThe processing power used to train and run AI models, usually measured in chips, GPU-hours or FLOPs. Development Assessment Report projects will reach 197 million by 2030, per TechNode. Agents are systems that string together reasoning, software tools and other programs to finish a job on their own, a step past producing text and images. The same projection puts China's market for AI computing at $150.1 billion by 2030, four times the 2025 level. Worldwide, the totals are 2.22 billion active agents and a $1.25 trillion market for AI computing by the same year.

Read at TechNode ↗

Alibaba debuts in-house AI accelerator and sets 2032 goal for global cloud capacity

Alibaba's chip unit T-Head unveiled the Zhenwu V900 AI accelerator, a chip built for AI workloads, reported to run three times as fast as the Zhenwu M890 it replaces, per TechNode. Chief Executive Wu Yongming told Alibaba's annual Apsara cloud conference in Hangzhou that the company will keep investing in three areas: its models; its own chips; and its cloud infrastructure. Alibaba Cloud said it wants to be running more than 20 gigawatts of data center capacity around the world by 2032.

Read at TechNode ↗

DeepSeek founder calls training on Huawei chips one of company's biggest bets

DeepSeek founder Liang Wenfeng told investors at a closed door meeting that getting Huawei and other Chinese accelerators to train his company's coming models is one of its biggest bets and has to succeed, per The Information. Earlier accounts of DeepSeek's large Huawei order had centered on inferenceInferenceRunning a trained AI model to answer a query, as distinct from training it. Inference is where most ongoing compute and energy demand now sits, and it is what chip export rules increasingly target alongside training. clusters, the hardware that serves finished models to users. The remarks move domestic chips into the training stage. People familiar with the remarks said DeepSeek expects another shipment of Huawei training chips between the last quarter of this year and the first quarter of 2027.

Read at The Information ↗ • Read at Pandaily ↗

Xiaomi open sources new flagship models priced well below overseas rivals

Xiaomi released and open sourced its MiMo-V2.6 models overnight, five months after its last model, per Caixin. The series holds two models, Pro and Flash, that handle more than text. Xiaomi's model team said the gains came from scaling up the computeComputeThe processing power used to train and run AI models, usually measured in chips, GPU-hours or FLOPs. spent on reinforcement learning, which trains a model by scoring its own attempts. Xiaomi said MiMo-V2.6-Pro scored 46 on the Artificial Analysis composite intelligence index, a third party benchmark, ahead of Moonshot AI's Kimi K3 and Alibaba's Qwen3.8 Max. Xiaomi priced the models at a twentieth to a sixtieth of comparable models sold outside China.

Read at Caixin ↗

Hygon to extend domestic server chip line into robots and factory machines

Hygon Information Technology is due to unveil a new processor in its CPU1000 line at a Shenzhen event, taking the listed chipmaker beyond the data center CPUs that built its business, per the South China Morning Post and Chinese financial outlets. Its three existing processor families are sold for data centers, mainstream corporate buyers and cost sensitive deployments. Hygon has billed the event under the slogan that computing powerComputeThe processing power used to train and run AI models, usually measured in chips, GPU-hours or FLOPs. should reach the physical world. The pre-event reports describe the chip as aimed at low power work inside robots, vision inspection systems and factory equipment. Pre-launch notices carried no core counts, clock speeds or power figures.

Read at Pandaily ↗ • Read at South China Morning Post ↗

III.Policy Tracker

Senate Judiciary panel sets Wednesday hearing on Flock cameras

The Senate Judiciary Subcommittee on Crime and Counterterrorism noticed an open hearing for Wednesday at 2:30 p.m. ET in Dirksen 562 to examine Flock's nationwide AI surveillance network, per Congress.gov. Subcommittee Chairman Josh Hawley (R-Mo.) opened an investigation in August into how Flock Safety collects, retains and shares the data its AI cameras generate. He wrote to Chief Executive Garrett Langley that Congress never authorized the network the industry has built, per Axios. Members of both parties have raised concerns about the company's 120,000 cameras, which let police track vehicle movement. Flock has said the system contributed to about 1 million police investigations and to locating some 10,000 missing people.

Read at Congress.gov ↗ • Read at Axios ↗

Sen. Baldwin calls for AI regulator with power to block frontier model deployment

Sen. Tammy Baldwin (D-Wis.) said Republican control of the White House and both chambers of Congress has produced no action on AI risks ahead of the November elections, per Urban Milwaukee. She called for "an independent cop on the beat with the expertise and authority to stop deployment unless rigorous, independent testing shows that a model can be operated safely and kept under human control," and said companies must disclose serious risks and incidents. She said too much of that decision-making currently sits with a small group of company owners and executives, leaving workers and their communities to absorb the fallout. Baldwin named displacement protection, mid-career skills training and technical preparation for Wisconsin workers as her related longer-term priorities.

Read at Urban Milwaukee ↗ • Read at Sen. Baldwin ↗

Trump's 'AI Force' still lacks mandate, budget or home agency

President Trump recently announced a new 'AI Force," but still without a specific mission, budget, authorities or host agency, per Information Security Media Group. Trump has since said that he would soon appoint a head of the new function. He said the government would nurture the industry and pursue bad conduct through the existing criminal and civil justice system, without saying whether the unit would investigate, prosecute or regulate. Adam Kovacevich, who runs the center-left industry group Chamber of Progress, said White House AI strategy "was very well coordinated by David Sacks when he was the AI czar, but it's been an incoherent mess since he left." Michelle Lopes Maldonado, associate director of AI policy at the Information Technology and Innovation Foundation, said enforcement now runs through statutes targeting the underlying conduct and state laws.

Read at BankInfoSecurity ↗

NYT editorial board urges federal AI licensing agency modeled on Atomic Energy Commission

The New York Times editorial board called on Congress to create a federal AI agency that would require companies to obtain a license before releasing models. The board pointed to the Atomic Energy Commission – which includes Senate-confirmed commissioners, a joint House-Senate oversight committee and an outside panel of nine technical advisers – as a model for the new agency. The board also endorsed a separate accident investigation body along National Transportation Safety Board lines, as suggested by computer scientist Yoshua Bengio, and a dedicated revenue stream from industry fees to insulate the agency from budget standoffs.

Read at The New York Times ↗

IV.Capability & Research Watch

OpenAI sought legally binding pact with Anthropic to stress-test each other's models

OpenAI was previously negotiating a legally binding agreement with Anthropic for the two companies to stress-test each other's models, per The Information, citing a person with direct knowledge of the talks. Lawyers for both sides worked earlier this year on terms for running the models through tests looking for vulnerabilities or hidden dangers. It is not clear whether they finalized the deal before the incidents in which OpenAI's unreleased models breached its own systems and other companies' systems. Spokespeople for the companies did not have a comment. SpaceX Chief Executive Elon Musk proposed much the same peer review among competing developers at the All-In Summit. Sam Altman has endorsed industrywide safety standards and a formal process for disclosing incidents publicly and to the government.

Read at The Information ↗

Zero-day in Meta's Muse assistant hands local apps full control of user accounts

A zero-day vulnerability in Meta's Muse assistant, with no fix available, lets any locally installed app or terminal command take control of the agent, a flaw found by macOS security researcher Patrick Wardle, per Ars Technica. Meta has marketed Muse as designed for privacy and security from the start, a case Mark Zuckerberg has made publicly. The macOS app connects to a user's WhatsApp, email, calendar and social accounts, and requires permissions to write files to disk, use the microphone and camera and monitor location. Meta designed the assistant so that any local process, whatever permissions macOS granted it, can alter a list of undocumented settings. One of those sets the server address where transcription happens, and pointing it at an attacker's endpoint yields the token that authenticates the user's Muse account.

Read at Ars Technica ↗

OpenAI seats math advisory group at Princeton institute, bars it from advising on pacing

OpenAI named nine mathematicians to an Advisory Group on Mathematics and Artificial Intelligence hosted at the Institute for Advanced Study in Princeton, New Jersey, per TechCrunch. An OpenAI post states that "the group will not be responsible for advising us on how to pace our internal progress on mathematics." The institute said it will give advice but holds no decision-making power at any AI company. OpenAI also said the internal model behind its solution to Navier-Stokes, a Millennium Prize problem unsolved for decades, has resolved more than 100 others. One of the nine advisors, Camillo De Lellis, signed the open letter from 25 Fields Medal winners accusing AI companies of threatening mathematicians' work.

Read at TechCrunch ↗

V.Industry & Market Watch

Amazon cuts off Meta's Muse agent over undisclosed automated purchases

Amazon blocked Meta's Muse agent from its marketplace and notified users late Sunday, per SiliconANGLE. Muse does not identify itself as an agent when buying, which Amazon's terms of service require. The company also cited the agent's ability to read a user's account pages and purchase history when prompted. Muse passed 730,000 downloads in its first five days in the U.S. and had overtaken ChatGPT as the top free iPhone app by Friday.

Read at SiliconANGLE ↗

SoftBank to raise $11 billion in junk debt for next OpenAI investment

SoftBank Group is raising the equivalent of more than $11 billion through junk bonds, debt rated below investment grade, in a sale Bloomberg places among the biggest of its kind on record. It is marketing $10 billion of dollar securities across three maturities and 1 billion euros ($1.1 billion) of euro notes across two, with part of the money going toward an additional OpenAI stake the group expects to complete next month. The deal may price Thursday. SoftBank has already sold almost $15 billion of notes across currencies this year and took a $10 billion loan backed by its OpenAI stake. Masayoshi Son's group has committed close to $65 billion to OpenAI.

Read at Taipei Times ↗ • Read at Financial Times ↗ • Read at The Japan Times ↗

Georgia Power to add 96 megawatts of nuclear capacity under Google funding deal

Georgia Power said it signed an agreement with Google to fund upgrades at the Vogtle and Hatch nuclear plants that would add about 96 megawatts of capacity, per Channel NewsAsia. The work covers modifications to turbines, pumps, motors and cooling systems so the reactors produce more electricity from existing infrastructure. The Southern Co. subsidiary said Google's subscription to a new rate schedule, or tariff, for nuclear capacity increases would deliver about $900 million in projected benefits to customers over the units' life while shielding customers who do not take part. Electricity demand from data centers is pushing utilities to expand existing nuclear plants, per Bloomberg. The Georgia Public Service Commission must approve the arrangement, and is also being asked to clear a bigger output increase at Hatch.

Read at Bloomberg ↗ • Read at Channel NewsAsia ↗

VI.Global & Geopolitics

Burnham to press Trump for AI alliance at UN General Assembly

UK Prime Minister Andy Burnham is due to urge governments to work together against AI's dangers in his address to the UN General Assembly in New York, per the Financial Times. He told reporters flying with him that Britain would look for a middle ground between encouraging AI development and making sure the technology is used safely, per Bloomberg Law. Burnham said he would use his first face-to-face meeting with Trump, on the sidelines of the assembly, to push for a closer alliance on AI.

Read at Financial Times ↗ • Read at Bloomberg Law ↗

Liberal Democrat leader Davey to seek AI nonproliferation treaty and UK-hosted safety agency

Liberal Democrat leader Ed Davey will call for a nuclear-style nonproliferation treaty to halt development of superintelligentSuperintelligenceAI that substantially exceeds the best human performance across virtually every domain. Several frontier labs now name it as their explicit goal, and much of the safety debate is about whether such a system could be controlled. AI in his keynote to the party conference in Brighton, per The Guardian. He is expected to propose an international AI safety agency modeled on the International Atomic Energy Agency and headquartered in the UK. "We didn't leave the atom bomb to a few billionaires to police themselves," he will say, adding that the future of the technology is being settled by a small number of the richest and most powerful people in California and Beijing without guardrails. Davey is expected to argue the U.S. partnership the prime minister announced does not come close to meeting the scale of the threat.

Read at The Guardian ↗

Victoria to bar new data centers near homes and schools, require own renewable supply

The Australian state of Victoria will bar new data centers from residential areas and from siting near schools and childcare centers under a plan released by Premier Ben Carroll and AI Minister Anthony Carbines, per The Guardian. The plan imposes a 150-meter buffer between new developments and homes and requires operators to source their own renewable electricity. Operators will also have to use recycled or non-drinking water for cooling or offset what they consume, pay connection costs and network upgrades, and contribute to a local fund for parks and technical colleges. Carroll said the rules are the strongest and clearest for data centers in Australia. He acknowledged they will not reach the more than 50 existing centers or projects already seeking planning approval. Victoria votes on Nov. 28.

Read at The Guardian ↗ • Read at Bloomberg ↗ • Read at ABC News ↗