AI Regulation & Policy

Today's Stories

1. 1

   Mastercard, Visa, Ant pilot Know-Your-Agent for AI checkout

   Mastercard, Visa and Ant International unveiled a Know-Your-Agent initiative on Thursday via the Monetary Authority of Singapore-backed BuildFin.ai platform, connecting Mastercard Verifiable Intent, Visa's Trusted Agent Protocol and Ant International's Agentic Mobile Protocol. The goal: let merchants trust software shopping on a customer's behalf. The pact tackles who pays when an autonomous agent authorizes the wrong transaction — a question that only grows as AI agents bring more purchasing activity onto card networks like Mastercard, which handled 47.4 billion switched transactions last quarter and generated $9.28 billion in revenue. Shared identity standards may cut fraud and confusion, though adoption timetables, a revenue model and loss-allocation rules remain undefined.

   Whether the initiative moves from protocol-connection to a defined revenue model, adoption timetable and clear loss-allocation rules — the three missing pieces investors care about. Mastercard shares slipped about 0.4% to $565.095 in morning trading, and the stock trades 17% below its GF Value estimate of $680.81, a valuation gap that could hinge on such clarity.

2. 2

   Finnish MP's question draws government reply on superintelligent AI

   A citizen cold-emailed a Finnish MP, met him for lunch, and the MP submitted a written question four days later, drawing a government reply after four weeks. Finland's government said it seeks to "constructively promote the creation of international regulation on the development of superintelligent AI," a formal answer the citizen says stemmed from a cold email.

   The reply paired that commitment with an emphasis on balancing safety with innovation, so the test is whether Finland pushes actual restrictions. Watch the 21-day response rule that forced the government to answer.

3. 3

   Oracle's Project Jupiter renewables pledge leaves gas power unchanged

   Oracle proposed investing in 2 gigawatts of renewable energy for New Mexico as part of its $165 billion Project Jupiter data center for OpenAI, and pledged $1 million for carbon-capture research. The data center will still run on fuel cells using natural gas, so renewables won't directly cut its emissions; an earlier 2.2-gigawatt gas plan would have produced 14 million tons of CO2 a year.

   The project's fate hinges on two legal challenges before the New Mexico Supreme Court, which paused air quality permitting on August 23. Construction is 26 percent complete.

4. 4

   Meta Sued Over NameTag Face Data

   Parents and children in Illinois and California filed a proposed class action in Chicago alleging Meta extracted biometric data from their Facebook and Instagram photos without notice or consent to build NameTag and train Emu and Muse Image. The complaint cites a 2020 $650 million Illinois settlement and a 2024 $1.4 billion Texas payment over biometric data, and it seeks $5,000 per intentional violation under Illinois law.

   The case hinges on whether the court accepts Meta's position that nothing shipped to consumers, and on how far back the proposed class reaches, dating to September 4, 2021.

5. 5

   Garry Tan: let U.S. open-weight labs distill frontier AI

   Y Combinator CEO Garry Tan told CNBC he would do nothing about distillation, and told TechCrunch he wants smaller American open-weight labs to distill U.S. frontier models so open-weight options aren't only Chinese. His stance splits from Anthropic, which this week released its second report alleging Chinese labs run "illicit distillation attacks"; CEO Dario Amodei had publicly called for a regulatory crackdown.

   The test is whether Tan's "American distillation regime" idea gains traction against Amodei's call for enforcement; Tan says the real nightmare is one monolithic proprietary AI company.

6. 6

   Anthropic exit sparks AI fear aimed at wrong risks

   Researcher Jacob Coxon resigned from Anthropic this week, claiming a good chance its work leads to human extinction. The backlash conflated that with data-center energy costs, hacking, and job-market effects. The term "AI safety" has expanded from "extinction" to cover many issues since March 2024. Advocates like Eliezer Yudkowsky said then that using it was a bad idea — and the article says it still is.

   The test is whether lawmakers and regulators tackle each risk separately. Overbroad rules could solve nothing, or slow safer and more economically beneficial AI, the article argues.