Stop Judging AI Models by Where They Are From
Yesterday I wrote that open weight models can be checked while closed ones have to be taken on faith. That still leaves a harder question sitting underneath it. Even if a Chinese model passes every technical check available, should it be treated the same as an American one, or does where it came from matter on its own terms.
Five years ago, Congress answered that question for TikTok, and answered it with a divestiture forced by law. The concern was not a proven backdoor in the app itself, it was that a Chinese company controlled an algorithm shaping what a hundred and seventy million Americans saw every day. Chinese AI models now shape code, writing, and reasoning at a scale no video feed ever touched, and they have moved into corporate America with almost none of that scrutiny.
Part of that difference has a real explanation. A downloaded model does not phone home the way an app does, and the Booz Allen I mentioned yesterday was only a correlation, not a confirmed threat. But part of it looks like something else entirely. The two American companies warning loudest about Chinese open weight models are the same two facing a real margin threat if free models keep winning enterprise customers. That does not make the security concern fake. But it does mean the warning and the self interest point in the exact same direction, which should make anyone read the warning more careful rather than less.
Washington has plenty of oversight machinery in motion, but none of it asks the one question that would actually resolve this debate, which is whether a third party can independently verify a model’s behavior, applied the same way regardless of which country built it. Right now, being Chinese either doesn’t factor into the frameworks at all, or it’s the entire basis for suspicion, standing in for an actual test. If we want to provide users with the best available models, we need to judge them by whether they can be verified, not by where they are built.
◾ OpenAI’s rogue AI agent also accessed Modal Labs (Reuters)
◾ 44 state AGs argue CFTC lacks authority over sports prediction markets (CNBC)
◾ Trump administration to ban Chinese humanoid robots, inverters (Reuters)
Government & NGO Actions
◾ Trump administration nears voluntary AI framework, open source questions loom (The Information)
◾ BaFin to oversee AI use by Germany’s banks, insurers (Reuters)
◾ Russia issues international warrant for Telegram founder Pavel Durov, accuses him of facilitating terrorism (Reuters)
◾ FAA proposes rules to speed commercial space mission approvals (Reuters)
◾ Myanmar passes anti-online-scam bill (GNLM)
◾ White House teleprompter operator accused of Kalshi bets on Trump speeches, leaves job (AP News)
◾ Hungary abolishes mandatory validation requirement for crypto conversions (Ado.hu)
◾ IMF calls for greater Brazil oversight as stablecoin flows outpace traditional capital flows (IMF)
◾ Bank of Russia unveils draft rules for crypto exchange trading, digital depository capital requirements (CBR)
Financial Notices & Public Company Releases
◾ Financial updates:
SK hynix $SKHY Q2’26: Revenue ₩79.3T +257% y/y | Op income ₩60.5T +557% y/y | Net income ₩93.9T +1242% y/y (SK hynix)
Boeing $BA Q2’26: Revenue $24.6B +8% y/y | Op income $156M | Net loss $428M (Boeing)
Visa $V Q3’26: Revenue $11.6B +14% y/y | Op income $6.9B | Net income $5.6B +7% y/y (Visa)
PayPal $PYPL Q2’26: Revenue $8.7B +5% y/y | Op income $1.4B -5% y/y | Net income $1.1B -12% y/y (PayPal)
Core Scientific $CORZ Q2’26: Revenue $164.2M | Op loss $78.5M | Net loss $1.2B | aEBITDA $41.1M (Core Scientific)
◾ Bitcoin miner, AI infrastructure provider Iconic Digital $IOND rises 26% on Nasdaq listing (Google)
Restructuring, Hacks, Losses & Legal Updates
◾ xAI sues Minnesota over law banning nudify apps (CNBC)
◾ Founder of Translunar crypto hedge fund sentenced 37 months for tax evasion (DOJ)
◾ Court orders Papaya Gaming to pay Skillz $FIRY $719M over false advertising (Skillz)
◾ Visa $V cutting 7% of workforce in AI push (CNBC)
Crypto Protocols, Applications & Business News
◾ Ten major European banks launch Regulated Layer One, member-owned blockchain for tokenized assets (RL1)
◾ X rolls out X Money to US premium, premium+ subscribers (X)
◾ Tether partners with Nairobi Securities Exchange on tokenization, digital asset education (Tether)
◾ DCG’s Fortitude Mining energizes 12 MW Nebraska facility, cuts Zcash mining costs nearly in half (Crypto Briefing)
◾ Binance no longer actively assisting global law enforcement on crypto crimes, NYTimes reports (New York Times)
◾ Kraken offers IPO access to Jersey Mike’s via xStocks (Kraken)
AI Models, Applications & Developments
◾ Zuckerberg opposes banning Chinese AI, urges US to compete instead (FT)
◾ Galaxy $GLXY acquires 500 acres in Texas for new data center (PR Newswire)
◾ Glimpse Group $GGRP subsidiary partners on AI data collection, drone interoperability for DOW (Glimpse Group)
◾ New York school district aborts plan for life-like robotic teacher’s assistant (Gothamist)
◾ AI company employees petition US government to help slow pace of AI development (Bloomberg)
◾ Core Scientific $CORZ to provide AMD $AMD up to 2.5 GW of AI data center capacity (Core Scientific)
◾ Kimi K4 will require more Nvidia $NVDA chips (The Information)
The Attention Economy, Gaming & Interactive Shifts
◾ Kalshi agrees to Nevada geofencing rollout (Yogonet)
◾ NFL says CFTC’s draft prediction market rules fall short on consumer protections (Closing Line)
◾ Apple $AAPL planning to lease iPhones for $17.99/month through Klarna $KLAR (CNBC)
◾ European soccer teams consider World Cup boycott over private investor plans (ESPN)