日本語
2026-07-18 Evening edition
Evening edition

AI News Daily
2026-07-18

A retrospective edition: safety design, courtrooms and gigawatts — the AI race is now being fought off the leaderboard.

Four stories — corporate developments and model releases
02
Today's highlights
01
OpenAI sets out a safety policy for teenage users
Corporate developments — Published: 2026-07-17
02
Apple sues OpenAI over alleged trade secret theft
Corporate developments — Published: 2026-07-10
03
OpenAI makes GPT-5.6 generally available
Model release — Published: 2026-07-09
04
Anthropic expands its compute deal with Google and Broadcom
Corporate developments — Published: 2026-04-07
03
Story 01 — Corporate developments
OpenAI sets out a safety policy for teenage users
Published: 2026-07-17
  • OpenAI argues on its official blog that teenagers deserve safe access to AI, and announces stronger protections around it.
  • Age prediction is being rolled out, alongside parental controls for quiet hours, disabling voice mode and managing image generation.
  • Users are prompted to take breaks during long sessions.
WHY IT MATTERS

With regulatory pressure on minors' use of AI rising worldwide, a major platform moving first on voluntary safeguards can become the reference point for industry standards and national rules.

3
strands of protection: age prediction, parental controls, break prompts
Tier 1 source: OpenAI official web.
Source: https://openai.com/index/why-teens-deserve-access-safe-ai/
04
Story 02 — Corporate developments
Apple sues OpenAI over alleged trade secret theft
Published: 2026-07-10
  • Apple filed suit in the U.S. District Court for the Northern District of California against OpenAI and two former Apple employees.
  • The claim covers hardware-related trade secrets and product design information; the complaint notes that over 400 former Apple staff now work at OpenAI.
  • OpenAI rejected the claim, stating it has no interest in other companies' trade secrets.
WHY IT MATTERS

A courtroom fight over talent and technology between an AI leader and a hardware giant crystallises the tension between aggressive AI recruiting and intellectual property protection.

400+
former Apple employees now at OpenAI, per the complaint
Tier 2 sources: TechCrunch, Axios.
Sources: https://techcrunch.com/2026/07/10/apple-sues-openai-over-alleged-trade-secret-theft/ | https://www.axios.com/2026/07/10/apple-sues-openai-trade-secret-theft
05
Story 03 — Model release
OpenAI makes GPT-5.6 generally available
Published: 2026-07-09
  • The GPT-5.6 family — Sol, Terra and Luna — is now formally generally available.
  • Additional review by CAISI at the U.S. Department of Commerce lifted the government restriction, turning the late-June limited preview into a full rollout.
  • Availability arrived in stages across ChatGPT, Codex and the API.
WHY IT MATTERS

As the first major model release to ship after a government review, it may set the precedent for building government involvement into how frontier models reach the market.

3
models in the family: Sol, Terra and Luna
Tier 1 source: OpenAI official web.
Source: https://openai.com/index/gpt-5-6/
06
Story 04 — Corporate developments
Anthropic expands its compute deal with Google and Broadcom
Published: 2026-04-07
  • Anthropic expanded its partnerships with Google and Broadcom, contracting next-generation TPUs at multiple-gigawatt scale.
  • The capacity is scheduled to come online from 2027 onward.
  • The deal is positioned as part of the USD 50 billion U.S. compute infrastructure investment announced in November 2025.
WHY IT MATTERS

The contest for compute is now measured in gigawatts, making these deals a key input for investment decisions across the power and semiconductor supply chains.

$50B
U.S. compute investment announced in November 2025
Tier 1 source: Anthropic official web; Tier 2: TechCrunch.
Sources: https://www.anthropic.com/news/google-broadcom-partnership-compute | https://techcrunch.com/2026/04/07/anthropic-compute-deal-google-broadcom-tpus/
07
Trend overview
Three axes of a more grounded contest

Advantage is still fluid

The frontier model race is intensifying, but stumbles are surfacing too — Google, for one, has delayed a model after missing performance targets. Technical leads are not locked in.

Talent, safety, compute

Litigation with an incumbent over people and IP, safety design for minors, and giant compute contracts: three axes that read as the industry moving to a more grounded stage of competition.

Governance hits the roadmap

A model released only after government review shows AI governance beginning to act directly on product roadmaps rather than from a distance.

08
Wrap-up
Three things to remember
WHAT TO WATCH NEXT

Whether rival labs publish equivalent minor-safety controls; whether the Apple suit ends in an injunction or a settlement; and whether government review becomes a standing step in frontier model launches.