日本語
Sources: [1] [2] [3] [4] [5]
2026-07-14 Morning edition
Morning edition

AI News Daily
2026-07-14

Usage limits and large model launches move together across all three major vendors, while transparency obligations turn into practical work.

Anthropic / OpenAI / Google DeepMind / EU / Japan
Today's highlights

Five stories

Story 01 — Developer tools

Claude Code extends its 50% weekly limit boost to July 19

Published: 2026-07-13
  • The campaign raising Claude Code weekly usage limits by 50% now runs to July 19, 23:59 PT, instead of ending July 13.
  • Pro, Max, Team and legacy seat-based Enterprise are eligible; it applies automatically with no action needed.
  • Covers CLI, IDE extensions, desktop and web. Limits for Claude on web, desktop, mobile and for Cowoker are unchanged.
Why it matters

Plan on limits returning to standard after July 19. Free and usage-based Enterprise plans are not covered.

50%
higher weekly usage limits, through July 19 at 23:59 PT
Source: https://www.anthropic.com/news/higher-limits-spacex / https://www.helpnetsecurity.com/2026/07/13/claude-code-weekly-limits-promotion-extended/
Story 02 — Developer tools

Codex shifts to token billing, credits extend past the cap

Published: 2026-07-13 (pricing structure revised 2026-04-02)
  • Codex billing has moved from per-message to API token usage.
  • ChatGPT Plus has a five-hour rolling limit per GPT-5.6 model: Sol 15 to 90, Terra 20 to 110, Luna 50 to 280 message equivalents, plus a weekly cap. Pro 5x and 20x raise those limits 5x and 20x.
  • After hitting a limit, users buy credits and keep working.
Why it matters

Claude Code lifts the ceiling; Codex charges past it. On a fixed budget, check whether auto top-up is on.

$20
per month for ChatGPT Plus, the entry tier with rolling-window limits
Source: https://help.openai.com/en/articles/20001106-codex-rate-card / https://developers.openai.com/codex/pricing
Story 03 — Model release

Gemini 3.5 Pro delayed to July 17, architecture rebuilt

Published: 2026-07-13
  • Google DeepMind pushed the launch to July 17 and discarded the older 2.5 Pro architecture for a full redesign.
  • A 2 million token context window, a Deep Think reasoning layer, and autonomous workflow capabilities.
  • Positioned against GPT-5.6 and Fable 5.
Why it matters

The delay reads as quality over schedule. Wait for post-release operational benchmarks before migrating.

2M
token context window in the rebuilt Gemini 3.5 Pro
Source: https://finance.biggo.com/news/6f0c6bb2-795f-4c57-9d09-6db691d7638a
Story 04 — Regulation and policy

EU weighs the Transparency Code ahead of August

Published: 2026-07-13
  • Transparency rules for AI-generated content are expected to apply in August 2026.
  • The EU is assessing the Transparency Code as an effective route to implementation.
  • Disclosure, labelling and accountability for generative AI content become practical corporate obligations.
Why it matters

Anyone supplying AI-generated content into the EU likely needs labelling in place before August. Japanese firms with EU-facing business need it too.

August 2026
when the AI-generated content transparency rules are set to apply
Source: https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
Story 05 — Japan

AI agents become an execution layer, not a chat tool

Published: 2026-07-14
  • The Japanese debate has shifted from whether to adopt agents to which work they are trusted with, where humans approve, and which logs are kept.
  • Agents are becoming an execution layer across web production, marketing, sales, accounting, internal IT and security.
Why it matters

Permission scope and log management are becoming standard design questions, for individuals and small businesses as much as for enterprises.

6
business areas named as adopting agents as an execution layer
Source: https://eguweb.jp/ai/81190/
Trend overview

Two clocks: capability and accountability

Limits and launches move at once

Anthropic, OpenAI and Google are all adjusting usage limits and shipping large models at nearly the same time. Developers now have to plan on two axes: how allowance is consumed, and when to switch models.

Transparency and governance

On the regulatory side, transparency and governance are shared keywords in both the EU and the UN. Practical accountability work is running in parallel with the performance race, not after it.

Wrap-up

Three things to remember

01 — Dated headroom

The Claude Code 50% boost ends July 19 at 23:59 PT. Treat it as a window, not a new baseline.

02 — Two limit models

Claude Code raises the ceiling; Codex sells credits past it. Compare consumption mechanics, not headline prices.

03 — Accountability is operational

EU labelling before August, and permission scope plus logging for agents. Both are build work, not policy reading.

What to watch next: limits reverting after July 19, real-world benchmarks for Gemini 3.5 Pro after the July 17 launch, and the final shape of the EU transparency obligation in August.