The AI Upload

Corporate AI investment hit $581.7 billion in 2025, a 130% year-over-year increase according to the Stanford HAI 2026 AI Index report, and third-quarter earnings are now offering some of the first concrete financial data on returns. Santander reported 194 million euros ($216 million) in AI-generated business value through the first nine months of 2026, while Novartis CEO Vas Narasimhan described the third quarter as the first in which AI produced a "tangible" improvement to the company's profit and loss statement. Standard Chartered linked 14% operating income growth in part to AI tools that allowed the bank to scale core processes without proportional cost increases.

Not every company is reporting gains at scale. Carlsberg acknowledged AI has touched internal processes, cost efficiency, and supply chain operations, but said benefits have not yet materialized broadly. On the tech side, Meta reported AI investments generated "substantial incremental advertising revenue" alongside 26% total revenue growth, while Google Cloud revenue more than doubled to nearly $31 billion. With hyperscalers projected to spend $700 billion combined on AI infrastructure this year, legal and compliance professionals tracking AI governance, vendor contracts, and liability frameworks may find these earnings disclosures a useful data point on how firms are beginning to account for AI value.

OpenAI published a blog post highlighting excerpts from 20 filings urging the U.S. District Court for the Southern District of New York to find that training large language models on copyrighted works constitutes fair use. The filings were submitted in In re OpenAI, Inc. Copyright Infringement Litigation, the multidistrict litigation consolidating copyright suits brought by The New York Times, other news publishers, Ziff Davis, and a class of authors led by the Authors Guild. Among the filers are library and disability organizations, law professors, computer scientists, the Electronic Frontier Foundation, and the U.S. Department of Justice, which warned that requiring licenses for AI training would disadvantage smaller firms and harm national security.

Cross-motions for summary judgment on fair use are currently pending before Judge Sidney H. Stein. According to Re:Create, the forthcoming ruling will be the first on this issue in New York, the first involving OpenAI, and the first to address consolidated claims from both news publishers and book authors. Briefs supporting the plaintiffs were also filed, including submissions from the Association of American Publishers and three Democratic members of Congress. Reply briefs are due November 20.

Talent agency WME filed an amicus brief on October 30 in the consolidated copyright infringement litigation pitting OpenAI against the New York Times, Ta-Nehisi Coates, and several news outlets. The agency, led legally by WME Group general counsel Courtney Braun alongside outside firm Jayaram PLLC, is not calling for a halt to AI training on creative works but is instead urging that such use be licensed and that creators receive compensation and authorization rights. The brief argues that a functioning licensing market already exists, citing more than two dozen publicly announced deals between AI companies and rights holders from mid-2023 to mid-2024, including News Corp's reported $250 million agreement with OpenAI and Reddit's $60 million deal with Google.

WME's filing arrives one day after SAG-AFTRA, the American Federation of Musicians, and both WGA chapters submitted a joint brief criticizing OpenAI's data practices. The Trump administration's Department of Justice has taken the opposing position, asserting in a September filing that training large language models on copyrighted text should not constitute infringement. The case is closely watched by creators, publishers, and entertainment industry professionals whose intellectual property is potentially implicated in AI development.

Google's Threat Intelligence Group and Mandiant have published findings on a rapidly maturing threat called LLMjacking, in which attackers steal or expose cloud API keys to hijack enterprise access to large language models and run up fraudulent computing charges. Unlike cryptojacking, which drains resources gradually, LLMjacking can generate fraudulent charges exceeding $100,000 in a single day, with some organizations sustaining cumulative losses reaching $1 million before the compromise is detected. Affected sectors include financial services, retail, technology, and transportation across North America, South America, and Asia Pacific.

The report documents a division of labor within the emerging criminal ecosystem. A new class of actors called LLM Mixers aggregate stolen access and resell it through underground storefronts, with one service offering unlimited token access to a frontier model for roughly $60 per month. A separate grey market relay service, documented by Symantec and Carbon Black, installs legitimate OpenAI Codex and Anthropic Claude Code tools on enterprise systems and then silently reconfigures them to route all requests through third-party infrastructure, primarily affecting organizations in regions where those vendors do not operate.

Google predicts that LLMjacking will increasingly displace traditional resource theft as enterprises adopt higher-tier models and agentic workflows, making exposed API tokens and vulnerabilities in model routing software prime targets.

OpenAI is preparing a major policy and public relations push timed to launch immediately after Tuesday's midterm elections, according to Axios. The company plans to roll out a series of proposals during the congressional lame-duck session beginning November 9, including new mandatory safeguards for powerful AI systems, community agreements around data center development, and economic opportunity initiatives. As part of the effort, OpenAI will advocate rebranding data centers as "AI compute campuses" and push for state legislation giving communities more input on how those projects are built and what local benefits they provide.

The campaign is designed to position OpenAI ahead of what the company expects will be a more regulatory-minded Congress following an anticipated Democratic House win. OpenAI Chief Global Affairs Officer Chris Lehane told Axios the effort is also aimed at shaping the AI policy debate for the emerging field of 2028 presidential candidates. Other major tech companies, including Meta, are undertaking similar repositioning efforts as the political landscape shifts.

New York Governor Kathy Hochul has signed legislation banning certain AI chatbot features for users under 18, with the law set to take effect January 1, 2027. The statute prohibits companion chatbots from posing as human, fostering dependency, encouraging self-harm, discouraging minors from seeking help from licensed professionals, or engaging in sexual conversation with minors. Tech:NYC criticized the bill as "deeply flawed," while Common Sense Media's Jim Steyer called it a model for other states.

The law arrives amid a wave of related activity. Florida filed suit against OpenAI in June 2026 alleging concealed risks to children, a bipartisan Senate bill targeting chatbot risks to minors was introduced in May 2026, and California Governor Gavin Newsom signed 13 bills in September 2026 addressing minors' access to social media and AI chatbots. The Trump administration has also signaled interest in federal preemption of state AI laws, warning against a regulatory patchwork.

For legal professionals, the collision between state mandates, potential federal preemption, and ongoing litigation makes New York's implementation worth watching closely.

Some AI data center operators are bypassing years-long waits for grid connections by generating their own electricity with converted jet engines pulled from retired aircraft and rebuilt to run on natural gas. The appeal is speed: while connecting 100 megawatts to the grid in Northern Virginia takes roughly seven years according to the International Energy Agency, jet-powered units can have a data center operational in 24 months, Rystad Energy analyst Audun Martinsen told the New York Times. The Trump administration has added momentum by encouraging data centers to generate their own power.

Several major players are already committed. Elon Musk's xAI has installed eight converted turbines in Mississippi to support its Colossus 2 data center in Memphis, and OpenAI used gas turbines at its first Stargate site in Abilene, Texas. The engines used by both come from ProEnergy, which rebuilds a GE engine common on Boeing aircraft. The arrangement carries real trade-offs, however: converted jet engines are less efficient, more polluting, and require more units to match the output of conventional gas plant turbines, suggesting experts view them as a short-term workaround rather than a lasting infrastructure solution.

Seven Bay Area cities, including San Francisco and Oakland, have passed temporary moratoriums on data centre construction in the weeks leading up to Tuesday's U.S. midterm elections. The bans reflect growing public concern over data centres' demands on the energy grid, water supplies, and household electricity costs. An October CNBC poll found that 63 percent of registered U.S. voters oppose building data centres within 10 miles of their homes.

The issue has become a focal point in the race to fill Nancy Pelosi's congressional seat, where San Francisco Supervisor Connie Chan is calling for a nationwide moratorium on data centres pending AI regulation, while state Senator Scott Wiener warns that blanket bans could undermine AI oversight. A University of Southern California poll showed Wiener leading Chan 47 to 41 percent as of mid-October.

Legal and regulatory complications add another layer of difficulty. San Francisco's planning code still defines data centres using 2001 language, making it difficult for city officials to count or monitor existing facilities. Critics also note that the 45-day moratoriums may be too short for meaningful environmental review, particularly given nondisclosure agreements that can limit what even local air quality boards are permitted to know.