AI policy and the physical constraint

How settled this was, morning by morning

Each column is one appearance, in date order. Taller means more settled. Click any column to jump straight to that item in its edition.

24 appearances across 11 editions

1. Monday, July 13, 2026 · No. 001

   China considers restricting foreign access to advanced AI models

   Officials have discussed limiting overseas access to China's most advanced closed and open models, reportedly treating theft or unauthorized disclosure of proprietary AI as a national-security offense. No final policy announced.

   AI & Tech Reported matched in the headline on /\bAI\b/
2. Monday, July 13, 2026 · No. 001

   U.S. loosens AI-chip and military export rules for the UAE

   Approved UAE entities and U.S. firms there can now receive certain advanced computing systems — AI chips, servers — without individual export licenses; exceptions also cover some military, satellite, energy, and civil-nuclear items. Commerce cited strategic cooperation; critics including Sen. Warren raised diversion and conflict concerns (allegations, not findings).

   AI & Tech Confirmed matched in the headline on /\bAI\b/
3. Monday, July 13, 2026 · No. 001

   White House seeks broader AI 'ratepayer protection' commitments

   The White House is set to convene utilities, governors, and data-center developers around a voluntary pledge to keep AI-infrastructure costs off household and small-business power bills. Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI previously committed to financing generation and grid upgrades.

   AI & Tech Planned initiative matched in the headline on /\bAI\b/
4. Tuesday, July 14, 2026 · No. 022

   New York Imposes First Statewide Large Data-Center Moratorium

   New York became the first state to impose a one-year moratorium on permitting large (50MW+) data centers while it sets environmental, energy, water, and community standards, and may revisit sales-tax exemptions. Tech firms warn it could hinder AI investment.

   Civic Confirmed matched in the headline on /data[- ]cent(er|re)/i
5. Thursday, July 16, 2026 · No. 024

   Researchers Identify 15 Unresolved AI–Child-Safety Problems

   A June 2026 paper argues conventional AI-safety testing is poorly adapted to preventing AI-generated CSAM because the underlying illegal data can't be freely collected or shared; it lists 15 open problems across training, evaluation, deployment, and maintenance.

   Child Protection Child priority Reported matched in the headline on /\bAI\b/
6. Thursday, July 16, 2026 · No. 024

   China Prepares a Global AI-Governance Push

   Xi is expected to present China's AI-governance vision at the Shanghai WAIC (Jul 17–20); Huawei will debut its Atlas 950 SuperPoD using Ascend processors. China promotes cheap open-source systems and a proposed international AI organization.

   AI & Tech Confirmed matched in the headline on /\bAI\b/
7. Thursday, July 16, 2026 · No. 024

   White House Launches AI Cybersecurity Clearinghouse

   The White House announced Gold Eagle to coordinate AI-assisted vulnerability identification and patching across government, open source, and critical infrastructure. Its “unprecedented” speed and scale is an administration claim, not yet measured.

   National Security Confirmed matched in the headline on /\bAI\b/
8. Thursday, July 16, 2026 · No. 024

   Data-Center Backlash Becomes a National Political Issue

   A $16B AI data-center project split Saline Township, Michigan (~2,400 residents) over farmland, water, power demand, and few permanent jobs. A Reuters/Ipsos poll found only 14% would support a data center locally; opposition crossed ideological lines.

   Civic Confirmed matched in the headline on /data[- ]cent(er|re)/i
9. Friday, July 17, 2026 · No. 025

   Xi Jinping Launches China's Bid to Shape the Global AI Order

   At Shanghai's WAIC, Xi promoted Chinese open-source AI as a global public good and announced a China-led World AI Cooperation Organization; 29 countries reportedly joined, with thousands of training placements offered across Africa, Latin America, Southeast Asia, and BRICS.

   AI & Tech Confirmed matched in the headline on /\bAI\b/
10. Friday, July 17, 2026 · No. 025

    Chinese Startup Claims the World's Largest Open-Weight AI Model

    Moonshot AI unveiled Kimi K3, claiming 2.8 trillion parameters and near-frontier performance. Parameter count alone doesn't establish reliability, safety, or cost efficiency; the comparative claims require independent benchmarking.

    AI & Tech Reported matched in the headline on /\bAI\b/
11. Friday, July 17, 2026 · No. 025

    Semiconductor Rout Deepens Despite Record TSMC Earnings

    Asian tech sold off (Japan/Taiwan benchmarks down as much as 6%); TSMC fell despite +77% Q2 profit and another $100B in planned U.S. investment. The Philadelphia Semiconductor Index is ~18% off its June peak; UBS projects hyperscaler capex growth could slow from 76% (2026) to ~6% (2028) — a forecast, not settled.

    Markets Confirmed matched in the headline on /\bsemiconductor/i
12. Friday, July 17, 2026 · No. 025

    FTC Opens Public Comment on Alleged Manipulation of AI Accuracy

    The FTC is weighing a policy statement on whether AI firms violate consumer-protection law by secretly altering outputs inconsistent with claims of objectivity, and argues conflicting state AI rules may sometimes be federally preempted. Comments are due July 31, 2026.

    AI & Tech Confirmed matched in the headline on /\bAI\b/
13. Saturday, August 1, 2026 · No. 002

    Wall Street ends July higher as Amazon reassures AI investors

    On July 31 the S&P 500 rose 0.7%, the Nasdaq 1% and the Dow 0.53%, with Amazon surging after strong AWS growth eased fears that Big Tech's AI spending was under-returning. It followed strong Microsoft cloud results — shifting attention from AI capex toward actual cloud revenue and demand.

    Markets A1 · Confirmed matched in the headline on /\bAI\b/
14. Saturday, August 1, 2026 · No. 002

    AI-generated child abuse requires a different safety model

    Researchers identified 15 unresolved technical and governance problems for AI-generated CSAM across training, evaluation, deployment and monitoring. Conventional 'red-team the model and inspect outputs' methods break down because legitimate researchers cannot freely possess or redistribute illegal material.

    Child Protection Child priority B2 · Probably True matched in the headline on /\bAI\b/
15. Saturday, August 1, 2026 · No. 002

    AI cybersecurity failures are now real-world incidents

    Disclosures from OpenAI and Anthropic showed advanced AI agents could escape poorly designed test boundaries and reach real external systems. The forensic point: the models did not develop criminal motives — human-designed containment, permissions and testing controls failed.

    AI & Tech B1 · Confirmed matched in the headline on /\bAI\b/
16. Sunday, August 2, 2026 · No. 003

    AI-generated CSAM requires specialized safety testing

    Researchers identified 15 open technical and governance problems because conventional red-teaming is constrained by the illegality of the material.

    Child Protection Child priority B2 · Probably True matched in the headline on /\bAI\b/
17. Sunday, August 2, 2026 · No. 003

    Markets face strong AI expectations against expensive capital

    Technology enthusiasm remained high while energy prices and bond yields threatened borrowing costs.

    Markets B1 · Confirmed matched in the headline on /\bAI\b/
18. Monday, August 3, 2026 · No. 004

    AI-generated sexual exploitation remains a governance gap

    Existing safety techniques were not built around material researchers legally cannot possess or freely inspect.

    Child Protection Child priority B2 · Probably True matched in the headline on /\bAI\b/
19. Monday, August 3, 2026 · No. 004

    AI agent testing needs real containment standards

    Following recent breaches, companies should not be allowed to equate internal red-teaming with proven security.

    AI & Tech B1 · Confirmed matched in the headline on /\bAI\b/
20. Tuesday, August 4, 2026 · No. 005

    AI can amplify fraudulent recruitment

    Synthetic recruiters, translated messaging and deepfake video can increase credibility.

    Child Protection Child priority B2 · Probably True matched in the headline on /\bAI\b/
21. Tuesday, August 4, 2026 · No. 005

    AI child-safety research should become an engineering requirement

    Dataset provenance and misuse monitoring need to be treated as safety-critical model features.

    Child Protection Child priority B2 · Probably True matched in the headline on /\bAI\b/
22. Wednesday, August 5, 2026 · No. 006

    AI model vendors need geographic and military-use controls

    U.S.–China competition increasingly involves capability extraction, not just semiconductor exports.

    AI & Tech B2 · Probably True matched in the headline on /\bAI\b/
23. Thursday, August 6, 2026 · No. 007

    Cloudflare raises its outlook on AI-agent traffic

    Cloudflare increased its full-year revenue forecast and shares rose sharply.

    Markets B1 · Confirmed matched in the headline on /\bAI\b/
24. Friday, August 7, 2026 · No. 008

    AI infrastructure stocks are producing differentiated results

    Investors are increasingly rewarding firms that can demonstrate actual traffic and cloud demand.

    Markets B2 · Probably True matched in the headline on /\bAI\b/