AI Regulation & Policy

Today's Stories

1. 1

   Microsoft–Chevron Power Deal: AI Competition Moves to Energy

   Microsoft signed a 20-year power purchase agreement with Chevron's subsidiary Energy Forge One LLC to develop Project Kilby, a West Texas facility providing dedicated electricity to a Microsoft data center. The project will deliver 2.67 gigawatts of capacity from GE Vernova turbines and Solar Turbines, supporting a multibillion-dollar campus expected to add about 2 gigawatts of global data center capacity over five to seven years. The deal marks a shift in AI competition from chips and cloud access to energy generation itself. Economists like Carl Shapiro and Fiona Scott Morton warn that controlling critical inputs, such as power, could raise rivals' costs and create chokepoints, echoing antitrust concerns from the Alcoa case about preempting future market capacity.

   Whether such dedicated power deals become a pattern of preemptive infrastructure commitments that lock up AI's essential inputs. Regulators, including the FTC and European and UK agencies, are already scrutinizing cloud and AI infrastructure control.

2. 2

   AAM Asks FTC to Probe AI Shopping Tools Over 'Made in USA'

   The Alliance for American Manufacturing (AAM) has asked the Federal Trade Commission (FTC) to investigate AI shopping tools on large online retail platforms. This follows a Columbia University Law School report finding that chatbots on Amazon and Walmart intentionally suppress "Made in USA" products in search results. The Columbia report says Amazon's chatbot cited its seller base, of which 40% to 50% is based in China, as a reason a "Made in USA" filter would redirect shoppers away from many products. AAM argues this could turn public concern about AI into a concrete marketplace harm, especially as December 2025 Morning Consult polling found nearly seven in 10 Americans concerned about AI's impact on domestic manufacturing jobs.

   AAM has sent the FTC a letter listing questions to explore, including whether AI tools obscure reliable country-of-origin info. AAM also supports Sen. Tammy Baldwin's COOL Online Act, which would require clear country-of-origin disclosures for online products.

3. 3

   Markey beats Moulton 65–35, wins Massachusetts Senate primary

   Ed Markey defeated Seth Moulton in the Massachusetts Democratic Senate primary on September 1, 2026, winning by a 65–35 margin and claiming another victory for the progressive wing of the party. Markey made AI regulation—including his detailed 'AI Accountability Agenda' with proposals like the No Robot Bosses Act and the Equal Tax Act—a central part of his campaign, while Moulton attacked him over AI knowledge; the outcome signals that voters backed the incumbent's substantive approach over the challenger's generational framing.

   Markey's agenda includes banning employers from primarily relying on automated systems for hiring, firing, or promotion, and taxing investment income at the same rate as labor income for individuals making more than $1 million.

4. 4

   Uber drivers file landmark European class action over AI pay algorithm

   Uber drivers from the UK, the Netherlands and other countries have filed a class action at Amsterdam's district court. They claim the AI-powered pay-setting system breaches data protection laws and pushes down earnings. The compensation claim could run into billions of dollars. Drivers say the opaque "black box" algorithm personalizes each ride's rate, often offering less for return trips after long jobs. Kola Oba from north London was offered £23 for the same job another driver got £27 for, and the pair suspected Uber used his history of cheap jobs against him. The lawsuit covers about 241,000 drivers across the EU and the UK.

   A 2025 University of Oxford study found substantial cuts in driver earnings after the algorithm was introduced, though Uber disputes its data. The Dutch data protection authority fined Uber €825m last month over automated account deactivation, which Uber is appealing.

5. 5

   OpenAI launches GPT-6 Astra, first to hit Critical cyber level

   OpenAI has released GPT-6 Astra, describing it as its most capable broadly deployed model and the first to reach the Critical level of cybersecurity capability under its Preparedness Framework. The Critical level signals a step up in AI's potential for both defense and misuse, which may lead to heightened scrutiny from regulators and security experts. OpenAI's own framework now classifies this model as critical, meaning it could pose severe risks if misused.

   How OpenAI and other developers handle the deployment of models at this level of capability, especially around safety protocols and potential restrictions on access.

6. 6

   AI needs electricity now; U.S. grid lags by years

   OpenAI, Google, and Meta can build AI infrastructure faster than the U.S. can build the electricity generation and transmission needed to power it, according to Rob Gramlich of Grid Strategies. The grid's slower pace means data centers may face years of delays or provisional, interruptible power connections. Data centers are projected to consume nearly 12% of all U.S. electricity by 2030, about six times their 2018 share, per Lawrence Berkeley National Lab. NERC's 2025 assessment forecasts summer peak demand in North America to grow by over 224 GW in the next 10 years, 69% above last year's projection, driven by new AI data centers.

   Kathryn Burke of Marsh predicts 50–60% of data center projects will be delayed and won't be operational within the one-to-two-year period companies hope for. U.S. power projects that came online in 2025 took a median of five years from grid connection request to commercial operation.