日本語
2026-07-10 Evening edition
EVENING EDITION
AI News Daily

AI News Daily
2026-07-10

Three labs shipped office-work agents in the same week, and both Illinois and the EU moved the compliance burden toward published risk plans and independent audits.

Today at a glance

Six stories

STORY 01Model release / Company moves

OpenAI merges ChatGPT and Codex, launches ChatGPT Work

Published: 2026-07-10 (announced 2026-07-09)
  • Codex is folded into the ChatGPT desktop app, putting Chat, Work and Codex in a single application.
  • ChatGPT Work, built on GPT-5.6, gathers information across apps, breaks a goal into steps and returns finished spreadsheets, slides and web apps.
  • Available on every plan including Free; the Atlas browser will be wound down in stages.
Why it matters

Folding a coding-specific tool into a general office agent means enterprise workflow automation can be attempted inside ChatGPT alone. Switching costs between rival tools fall, and the office-replacement agent race with Anthropic and Google begins in earnest.

Source: https://www.macrumors.com/2026/07/09/openai-chatgpt-work/
STORY 02Company moves

Anthropic expands Claude Cowork to mobile and web

Published: 2026-07-07 to 2026-07-09 (phased rollout)
  • Claude Cowork, an agent that keeps running tasks autonomously in the background, is rolling out to iOS, Android and the web, beta first for Max plan subscribers.
  • Chat and Cowork merge into one home tab, so work continues and can be monitored across devices.
  • Across about 1.2 million Cowork sessions, software development was only 8.7% of use.
Why it matters

Arriving days after ChatGPT Work, it shows the coding-agent war spreading to everyday non-engineering work. Real usage data confirms routine business process automation is the main market.

33.4%
Business-process work such as report writing and data aggregation, the largest category of Cowork sessions
Source: https://venturebeat.com/technology/anthropic-brings-claude-cowork-to-mobile-and-web-as-usage-data-shows-most-users-arent-coding
STORY 03Model release / Company moves

Meta ships Muse Spark 1.1 for coding and agents

Published: 2026-07-09
  • Meta Superintelligence Labs, led by Alexandr Wang, says the model matches GPT-5.5 and Claude Opus 4.8 on many agent evaluations.
  • A 1 million token context targets long tasks such as large-scale code migration and bug fixing; available same day via the Meta Model API and Meta AI.
  • Pricing is $1.25 per million input tokens and $4.25 per million output tokens. Wang says a stronger successor, Watermelon, is in training.
Why it matters

With Meta following OpenAI and Anthropic, every major player has put agentic coding at the center of its business. Enterprise buyers gain a materially wider shortlist.

1M tokens
Context window, sized for long-running migration and bug-fixing work
Source: https://www.cnbc.com/2026/07/09/meta-jumps-into-ai-coding-market-to-chase-anthropic-and-openai.html
STORY 04Regulation and policy (United States)

Illinois signs an AI safety law billed as the strongest in the US

Published: 2026-07-06 (effective 2028-01-01)
  • Governor Pritzker signed the AI Safety Measures Act (SB315).
  • First US provisions requiring frontier AI developers to publish catastrophic-risk plans and take an annual independent third-party audit.
  • Incidents must be reported to the state in as little as 24 hours. Effective January 2028.
Why it matters

With federal AI regulation stalled, states are tightening rules on their own. Companies operating in several states face compounding compliance work, and other states are expected to follow.

$3 million
Maximum penalty for a violation under SB315
Source: https://gov-pritzker-newsroom.prezly.com/gov-pritzker-signs-nation-leading-artificial-intelligence-safety-law
STORY 05Regulation and policy (EU)

EU publishes its action plan on cybersecurity and AI

Published: 2026-07-07
  • The European Commission published the EU Action Plan on Cybersecurity and Artificial Intelligence.
  • It sets a coordinated approach for member states, companies and public bodies facing the security and resilience challenges of state-of-the-art AI models.
  • A call for third-party evaluators to strengthen pre-market model assessment is planned, targeted to be operational in 2027.
Why it matters

The EU is filling in the AI Act’s operating detail step by step. With transparency rules due in August 2026, companies serving the EU must get evaluation and reporting arrangements in place now.

August 2026
AI Act transparency rules are due to take effect
Source: https://www.insideprivacy.com/artificial-intelligence/eu-ai-act-update-timeline-relief-targeted-simplification-and-new-prohibitions/
STORY 06Japan

neoAI allies with five critical-infrastructure firms at once

Published: 2026-07-09
  • neoAI, a generative AI startup out of the Matsuo Laboratory at the University of Tokyo, signed simultaneous capital and business alliances with five firms.
  • The partners are Aozora Bank, Johnan Shinkin Bank, Kyushu Electric Power, Kyoto Chuo Shinkin Bank and The Bank of Iwate.
  • With the Japan Post Bank alliance of May 2026, that brings the total to six partners.
Why it matters

Generative AI in finance and electric power is reaching beyond the largest institutions into regional banks, evidence that industry-specific adoption in Japan is broadening.

6 partners
Critical-infrastructure alliances in total, after the five signed together
Source: https://www.jiji.com/jc/article?k=000000051.000109048&g=prt
Trend overview

Three currents running through the day

01

Agents move into the office

OpenAI, Anthropic and Meta launched ChatGPT Work, Claude Cowork and Muse Spark 1.1 in the same window. Competition has expanded from coding assistance to white-collar work in general.

02

Regulation converges on audit

The US is acting state by state, starting with Illinois; the EU is writing operating detail beneath its existing AI Act. Both land on accountability and audit duties for model developers.

03

Japan adopts beyond the giants

Generative AI partnerships are reaching regional financial institutions and infrastructure operators, widening the base for industry-specific deployment.

Wrap-up

Three things to remember

What to watch next

The EU AI Act transparency rules due in August 2026 and the EU third-party evaluation capacity targeted for 2027; the January 2028 start of Illinois SB315 and whether other states copy it; and Meta’s next model, Watermelon, now in training.