AI Daily Briefing

Monday, July 06, 2026

Monday, July 06, 2026 (Asia/Seoul basis)

To: Attorney Kim Kyungjin

It was a quiet cycle over the weekend. The major announcements were concentrated last Thursday and Friday, and this briefing gathers developments not yet covered, along with regulations that are about to take effect. For each item, the publication date is listed as originally published.

1. Crusoe is in talks to raise $3 billion. Its valuation would jump to $30 billion, three times last October's level. It is the company supplying data centers to Meta, Oracle, and OpenAI. (Bloomberg, July 2)

2. The White House is finalizing a voluntary standard for frontier model releases. Anthropic, Google, Microsoft, Amazon, and the NSA are setting benchmarks, review periods, and access rules. Reports suggest an announcement may come this week. (Reuters, July 2 [follow-up])

3. Chinese large models have outpaced the United States in global weekly call volume for several consecutive weeks. Alibaba's Qwen3.6 Plus ranks first on OpenRouter's weekly chart. Open-weight leadership is hardening. (Securities Times-affiliated tally, late June to early July [follow-up])

• China's open-weight camp has locked in usage momentum. Qwen3.6 Plus ranked first on OpenRouter with 4.6 trillion weekly tokens, while Meituan's LongCat-2.0 and Zhipu's GLM-5.2 held top positions. The center of gravity in frontier AI is moving from closed U.S. systems to open Chinese ones. (Late June to early July tally [follow-up])

• The three U.S. frontier leaders are splitting at the release gate. Anthropic's Claude Sonnet 5 has already become the default model for free and Pro users. OpenAI's GPT-5.6 Sol remains tied to a government pre-approval gate, with general release expected in mid-July. Google's Gemini 3.5 Pro is waiting for a July general release due to quality tuning, and its context window is 2 million tokens. (TechTimes, TechCrunch, June 29 to July 2)

• The EU AI Act's enforcement authority over general-purpose AI (GPAI), along with content transparency and watermarking obligations, takes effect on August 2. As of today, 27 days remain. From that date, the European Commission may request information, evaluate models, order recalls or market restrictions, and impose fines. Violations are subject to the greater of 15 million euros or 3% of global revenue. (Latham & Watkins, early July [follow-up])

• China's Interim Measures for AI Anthropomorphic Interaction Services take effect on July 15. Nine days remain. They require dynamic identity disclosure, two-hour usage alerts, emergency response for extreme emotional states, and one-click termination for virtual companions and emotionally oriented chatbots. This is the first regulatory framework aimed at AI companions. (CAC [follow-up])

• The U.S. voluntary frontier model standard is near announcement. Rooted in President Trump's June 2 executive order, it is a voluntary framework under which developers grant the government access for up to 30 days before public release. It expressly states that this is not a mandatory licensing regime. (White House, Reuters, July 2 [follow-up])

• The FTC's request for comment on its policy statement on "AI accuracy" remains open until July 31. The core argument is that Colorado-style state laws that compel changes to AI outputs are preempted to the extent they conflict with federal regulation. The statement comes at a moment when the FTC's independence itself is under strain, after the Supreme Court last month overturned the Humphrey precedent and broadly recognized presidential authority to remove FTC commissioners. (FTC, July 1 [follow-up])

• Anthropic opened the Claude Science beta and, alongside it, disclosed an internal drug discovery program aimed at neglected diseases. It is a research workbench equipped with more than 60 tools for genomics, single-cell analysis, proteomics, and structural biology, and it leaves a reproducible audit trail. This is not a new model but a workflow product. (Anthropic, June 30 to July 5 [follow-up])

• Claude Code has formally released Chrome browser control as a standard feature. Background agents can handle automatic commit and push operations and draft pull requests, and a /dataviz skill has been added for chart design. (Anthropic Changelog, early July [follow-up])

• Crusoe is in talks to raise $3 billion, about KRW 4 trillion. Its valuation would reach $30 billion, triple last October's $10 billion. To date, it has raised a cumulative $2.77 billion and builds data centers for Microsoft, Oracle, and OpenAI. Capital is pouring into AI infrastructure. (Bloomberg, SiliconANGLE, July 2 to 3)

• According to Crunchbase, global venture investment reached $510 billion in the first half of 2026. That surpassed the full-year 2025 total of $440 billion in just six months. Of this year's AI investment, 88%, or $319 billion, flowed to U.S.-headquartered companies, with a large share concentrated in OpenAI and Anthropic. (Crunchbase, July 1)

• Technology: Claude Sonnet 5 is now the default model for free and Pro users. GPT-5.6 Sol is expected in mid-July, subject to a government approval gate. Gemini 3.5 Pro is awaiting general release in July.

• Law / Regulation: Announcement of the voluntary frontier release standard is imminent. The FTC's comment period on its AI accuracy policy statement closes July 31.

• Products / Use: Claude Science beta and its drug discovery program. Claude Code's formal Chrome control release.

• Investment: Crusoe's $3 billion funding talks. Of the $510 billion in global first-half venture investment, 88% of AI capital went to the United States.

• Technology: Weekly large-model call volume has exceeded the United States for several consecutive weeks. Qwen3.6 Plus ranks first on OpenRouter's weekly chart. GLM-5.2 and LongCat-2.0 remain leading open-weight models.

• Law / Regulation: The Interim Measures for AI Anthropomorphic Interaction Services take effect on July 15. This is the first regulation aimed at companions.

• Products / Use: DeepSeek V4's official release is signaled for mid-July. It will be the first to introduce peak and off-peak pricing by time of day.

• Investment: AI financing in the first half of 2026 exceeded RMB 300 billion. Funding concentrated in large models and embedded intelligence.

If you save repeated review work as a single "skill," variance in legal practice drops.

1. Write the review rules in one file. Set which clauses to check first, how to mark risk, and what format to use for the result.

2. Register those rules in Claude or ChatGPT as a skill, a saved workflow. From then on, you only need to provide the contract, and the same procedure will assist you.

3. For major matters, let the model ask back about issues, exceptions, and transaction background before it starts review. That makes the result more stable.

4. Use one model to map the issues, and another to draft clause-by-clause revisions. The intermediate output becomes the next input as is.

Without this habit, you waste time rewriting instructions for each recurring contract type, and review standards start to drift from person to person. (Artificial Corner, GenAI Unplugged, July 2026)

Three things overlap today. Money is flowing not to models but to power and data centers, while rules are hardening in different directions: U.S. voluntary standards, Europe's mandatory norms, and China's companion regulation. Usage has already tilted toward Chinese open-weight models. While the very best closed models remain tied to government approval gates, cheaper and more open models are widening their real-world user base. In practice, the safer course is to design the review process so that people retain control while models can be swapped in and out. The regulatory calendar also needs to be watched in advance. Europe on August 2 and China on July 15 are close.

The key is not how difficult it is, but whether it solves the problem. That is planning in the end. The real moat is not code, but experience, trust, and data.

Crusoe funding talks: https://siliconangle.com/2026/07/03/ai-data-center-builder-crusoe-reportedly-raising-3b-30b-valuation/

White House frontier voluntary standard: https://tradersunion.com/news/financial-news/show/2550448-white-house-ai-model-standards/

Trump June 2 executive order: https://www.whitehouse.gov/presidential-actions/2026/06/promoting-advanced-artificial-intelligence-innovation-and-security/

FTC AI accuracy policy statement: https://www.ftc.gov/news-events/news/press-releases/2026/07/ftc-seeks-public-comment-policy-statement-addressing-ai-accuracy

Supreme Court overturns Humphrey precedent: https://www.cnbc.com/2026/06/29/supreme-court-trump-slaughter-ftc.html

EU AI Act GPAI enforcement authority: https://www.lw.com/en/insights/eu-ai-act-gpai-model-obligations-in-force-and-final-gpai-code-of-practice-in-place

Claude Science: https://www.marktechpost.com/2026/07/04/anthropic-launches-claude-science-beta/

$510 billion in global first-half venture investment: https://news.crunchbase.com/venture/global-startup-exits-ipo-ma-soar-ai-q2-h1-2026/

DeepSeek V4 official release in mid-July: https://technode.com/2026/06/30/deepseek-to-launch-v4-in-mid-july-with-new-peak-time-api-pricing/

Using AI workflows: https://artificialcorner.com/p/claude-levels

(If you want to open multiple web links in new tabs at once, the Chrome extension Grabbit is convenient. You can right-drag to select several links and open them in new tabs, or hold the Z key and left-drag to copy only the link addresses.)

US AI frontier model release standard July 2026, White House voluntary AI standard Anthropic Google, FTC AI accuracy policy statement July 2026, EU AI Act GPAI enforcement August 2 2026, Crusoe $3 billion funding valuation July 2026, Crunchbase H1 2026 AI venture funding, 中国 大模型 调用量 周榜 Qwen July 2026, 中国 AI 拟人化互动服务 7月15日, DeepSeek V4 mid-July peak pricing, Claude Science drug discovery July 2026, ChatGPT Claude reusable workflow skill 2026

Generated at: July 06, 2026 06:00 KST