日本語
2026-07-12 Morning edition
AI NEWS DAILY

2026-07-12

MORNING EDITION

Government moved to the centre of AI this week: a frontier model shipped only after a federal safety review, the Federal Reserve gave AI a seat in its policy overhaul, and China set an enforcement date for the first rules on emotional AI companions — while capital kept pouring into the chips underneath it all.

01

Today's highlights

1
Apple sues OpenAI over alleged trade secret theft
Corporate · Published: 2026-07-10
2
OpenAI opens the GPT-5.6 family (Sol, Terra, Luna) to everyone
Model release · Published: 2026-07-09
3
SK Hynix climbs 13% on its Nasdaq debut
Corporate · Published: 2026-07-10
4
Federal Reserve sets up a task force on AI, jobs and productivity
Regulation and policy · Published: 2026-07-09
5
China's rules for anthropomorphic AI interaction services take effect
Regulation and policy · Published: 2026-07-15 (in force; announced 2026-04-10)
6
Kawasaki, FANUC and Yaskawa join forces on a physical AI dataset
Japan · Published: 2026-07-02 (GENIAC selection); 2026-07-08 (coverage)
02

Apple sues OpenAI over alleged trade secret theft

Published: 2026-07-10 · Corporate
  • Apple filed suit in the U.S. District Court for the Northern District of California, alleging theft of trade secrets.
  • A 24-year Apple veteran who became OpenAI's hardware lead is alleged to have used recruiting interviews to get candidates to bring in confidential parts and information on unreleased iPhone and Apple Watch work.
  • Apple says the taking occurred at every level, from technical staff up to the chief hardware officer.
24 years
Time the departing employee spent at Apple
Why it matters

As AI companies move into hardware, poaching disputes become IP litigation — a new legal risk between AI firms and incumbent tech giants.

Source: https://www.cnbc.com/2026/07/10/apple-openai-lawsuit-trade-secrets.html
03

OpenAI opens the GPT-5.6 family to everyone

Published: 2026-07-09 · Model release
  • GPT-5.6 shipped on July 9 in three tiers: Sol (flagship), Terra (balanced) and Luna (fast, low cost). It becomes the new default model in ChatGPT.
  • Release followed a U.S. Department of Commerce safety review, after a limited preview period held at the government's request.
  • API pricing per 1M tokens: Sol $5 in, $30 out; Terra $2.50 in, $15 out; Luna $1 in, $6 out.
3 models
Sol, Terra and Luna in one release
Why it matters

A government safety review before a frontier launch is an unusual step, and a sign that state involvement in AI development is deepening.

Source: https://openai.com/index/previewing-gpt-5-6-sol/
04

SK Hynix climbs 13% on its Nasdaq debut

Published: 2026-07-10 · Corporate
  • South Korea's SK Hynix listed on Nasdaq as an American Depositary Receipt and closed day one at $168.01, about 13% above the offer price.
  • The offering raised roughly $26.5 billion — the largest U.S. listing by a foreign company on record.
  • Demand reached about seven times the shares on offer.
13%
First-day close above the offer price
Why it matters

With high-performance AI memory in short supply, capital is accelerating into the semiconductor supply chain that carries AI infrastructure.

Source: https://www.cnbc.com/2026/07/10/sk-hynix-skhy-stock-nasdaq.html
05

Federal Reserve task force to study AI, jobs and productivity

Published: 2026-07-09 · Regulation and policy
  • New Fed Chair Kevin Warsh announced five external task forces as part of a comprehensive review of monetary policy.
  • a16z co-founder Marc Andreessen becomes co-chair of the "Productivity and Employment" group, examining how AI and other new technologies affect productivity, employment and growth.
  • Concrete recommendations are expected before the end of the year.
5 task forces
External groups in the Fed's policy review
Why it matters

A central bank formally treating AI's employment impact as a monetary policy question means AI's macroeconomic effects are now a top-tier policy agenda item.

Source: https://www.axios.com/2026/07/09/fed-warsh-task-force-andreessen
06

China's anthropomorphic AI rules take effect July 15

Published: 2026-07-15 (date in force; announced 2026-04-10) · Regulation and policy
  • Five authorities, including the Cyberspace Administration of China, bring interim measures on emotionally anthropomorphic AI conversation services into force on July 15.
  • Virtual-partner and virtual-family services for minors are banned, content suggesting self-harm is prohibited, and a safety assessment filing is required at 100,000 monthly active users.
  • ByteDance and Alibaba are already disabling the affected features ahead of the deadline.
100,000
Monthly active users triggering a safety filing
Why it matters

The first comprehensive regulation of the fast-growing emotional AI companion market, with the potential to shape product design in other countries and companies.

Source: https://www.hoganlovells.com/en/publications/chinas-interim-measures-for-the-administration-of-anthropomorphic-ai-interaction-services
07

Kawasaki, FANUC and Yaskawa build a physical AI dataset

Published: 2026-07-02 (GENIAC selection); 2026-07-08 (coverage) · Japan
  • Kawasaki Heavy Industries, FANUC and Yaskawa Electric — three competing industrial robot makers — were selected for GENIAC, the generative AI support programme of the Ministry of Economy, Trade and Industry and NEDO.
  • They will jointly collect 5,000 hours of video, tactile and motion data from manufacturing floors for a "VTLA model" integrating vision, touch, language and motion.
  • Osaka University, ABEJA and tactile sensor start-up FingerVision are also taking part.
5,000 hours
Joint data collection from factory floors
Why it matters

Direct rivals sharing a data foundation is unusual, and shows Japan's industrial policy aiming to secure competitiveness in physical AI.

Source: https://response.jp/article/2026/07/08/413754.html
08

Trend overview

State and AI

Policy is inside the loop

A frontier release (OpenAI GPT-5.6), an employment policy review (the Federal Reserve) and rules on emotional AI services (China) all point the same way: AI is now tied more tightly than ever to governments and regulators.

Capital

Infrastructure money runs hot

Funds continue to flood AI infrastructure — semiconductors and memory. SK Hynix's successful IPO stands as a symbol that the AI investment boom is still under way.

Talent and data

Conflict, and cooperation

Friction over people and IP between AI firms and incumbent tech giants has surfaced (Apple vs OpenAI), exposing legal risk as a side effect of the race. In Japan, competing robot makers are instead building a shared data foundation for physical AI.

09

Wrap-up

What to watch next

OpenAI's response in the Apple case; whether the Commerce Department review becomes a standing requirement; how Chinese companion features settle after July 15; and how specific the Federal Reserve's year-end recommendations turn out to be.