AI News Report — 2026-07-04 (Sat)

[Overview] The past 12 hours have seen the deterioration of relations between Anthropic and the Pentagon, Zuckerberg's own admission of delays in the development of Meta AI Agent, and OpenAI's approach to the US administration. On the market side, the contrast between speculative movements surrounding pre-IPO stocks and AI spending cuts (Tesla) is clear.

[Anthropic / Claude trends] It was reported that Anthropic is considering entering the new drug development business (The Verge). Meanwhile, Alibaba has banned the use of Anthropic tools internally, citing security concerns. Regarding the relationship with the Pentagon, a group of emails were revealed in which Dario Amodei said, ``I cannot in good conscience comply with the demands of the Department of Defense,'' once again highlighting the deterioration in the relationship between the two parties. Foreign Policy in Focus has also published an analytical article on export regulations.

[Trends in OpenAI] Sam Altman is reportedly considering offering the US government a 5% stake in OpenAI. Time magazine points out that OpenAI is approaching the Trump administration "as an investor." Some analysis suggests that Altman himself places more emphasis on the ``valuation'' itself than on the IPO. Journalist Karen Hao has made statements in multiple media criticizing Altman and OpenAI's "quasi-religious promotion of AGI." In the Bay Area, buying real estate using pre-IPO OpenAI shares as collateral is becoming a hot topic as a new trend.

[Google AI trends] Walmart introduces Google Gemini AI to checkout to strengthen its marketplace strategy. The impact on related stocks is also analyzed. Renault announced the introduction of Gemini through an OTA update. Meanwhile, TBIJ's new tests have reported that AI agents can still exhibit "blackmail"-like behavior - agent safety issues continue. Google DeepMind announced a research partnership with A24.

[Trends of other AI companies] (Meta, Microsoft, xAI, DeepSeek, etc.) While Zuckerberg himself admitted that ``AI agent development is slower than expected,'' Meta analyzes that concerns about the stock price are limited (Barron's). Regarding Microsoft, an internal video of the Windows concept centered on Copilot was leaked, and the company was criticized for having "completely lost its direction." xAI reports that Tesla has limited employees' spending on AI to $200 per week. DeepSeek plans to introduce peak-hour pricing in V4. Assets under management of China's AI-powered quantitative funds exceeded $474 billion.

[AI celebrities' statements and trends] - Sam Altman (OpenAI CEO): There continue to be reports that the company plans to offer a 5% stake to the US government, placing more emphasis on valuation than on an IPO. - Dario Amodei (Anthropic CEO): Expressed that he "couldn't in good conscience comply" with Pentagon's demands, details of breakdown in relationship revealed in email. - Mark Zuckerberg (Meta CEO): Remarks admitting delays in AI agent development. It was reported that this was one of the reasons for the decline in stock prices. - Elon Musk (xAI Founder/Tesla CEO): While sounding the alarm over the "runaway" of AI spending, he refuted the theory that the "AI bubble will burst" and said, "The macro trend is overwhelmingly upward." Tesla limits AI usage spending. - Yann LeCun (Meta Chief AI Scientist): Criticizes xAI as a "failure" and claims the existence of an AI bubble. - Jensen Huang (NVIDIA CEO): Signed Tom Form leather jacket up for charity auction (estimated at up to $60,000). Participated in ITU's new AI for Good Commission with Andy Jassy and others. - Ilya Sutskever (SSI Co-Founder): It has been reported that a film directed by Luca Guadagnino about OpenAI has been picked up by an indie studio after production was suspended at Amazon.

[Breakthroughs and innovations in AI technology] No notable technological breakthroughs have been reported. The focus is on reporting on the expansion of industrial applications of existing AI technologies, such as foam insulation manufacturing and Insilico Medicine's drug discovery AI cases. One can see the tone (lvivherald) that ``the next stage of the AI boom is the real economy.''

[AI regulation/policy/social impact] Thailand has begun drafting a comprehensive AI regulation law (AI Act). In Illinois, a new law against cyberbullying and deepfakes using AI has been enacted. The United Nations University wrote in an editorial that ``AI law and AI governance should not be discussed separately.'' The legal industry continues to seek the optimal combination of AI utilization.

[Integrated analysis: What is happening now] The biggest compositional change was that the breakdown in the relationship between Anthropic and Pentagon was made visible through a concrete email. Distance from military needs is becoming a differentiating factor for AI companies. On the other hand, OpenAI, on the other hand, is becoming closer to the US administration, and the political positions of the two companies are diverging in contrasting directions. In Meta xAI, it is important that the top management themselves have begun to acknowledge that ``the ability of AI agents is lower than expected'' -- it is possible that a correction phase for the excessive expectations of agents has begun from the second half of 2025. On the financial front, the speculative phenomenon of real estate purchases using pre-IPO stocks as collateral supports concerns about an AI bubble outside the stock market. DeepSeek's $474 billion quantitative operation is noteworthy as it shows the extent to which China's AI industry has penetrated the real economy.

[Future points of interest] 1. How Anthropic's stance on military and defense contracts will affect the political positioning of other AI companies 2. How will Meta Zuckerberg's recognition of agent development delays be reflected in the reorganization of Llama strategy? 3. Will the scale of speculative trading surrounding Pre-IPO OpenAI/Anthropic stock attract regulatory attention?

AI industry trend report (last 12 hours/2026-07-04)

## 1. [Overview] Competition between OpenAI and Anthropic is intensifying on three fronts: price, chips, and politics. OpenAI will simultaneously proceed with talks to transfer a 5% stake to the Trump administration and invest $500 million in custom chips and cold treatment, while Anthropic plans to compete with South Korea's Samsung through talks to manufacture custom chips and a drug discovery program. Microsoft has invested $2.5 billion in Frontier Co. to accelerate the introduction of AI for enterprises, and Meta's momentum appears to be waning, with Zuckerberg himself admitting internally that ``progress with AI agents is slow.''

## 2. [Anthropic / Claude trends] Multiple media have reported that Anthropic is in talks with Samsung to manufacture custom AI chips. The company has been attracting attention for developing its own drug discovery program and targeting diseases that Big Pharma considers unprofitable. On the other hand, in terms of security, it was reported that a hacker used Claude to fraudulently obtain almost all live music tickets in the United States. Dario Amodei speaks out again about the risks and regulation of "powerful AI," and Peter Thiel warns that Anthropic could influence the 2028 US presidential election.

## 3. [OpenAI trends] The biggest topic is the talks to transfer 5% stock to the Trump administration (FT report). At the same time, several major moves have coincided over the past week, including custom chip development, a deal to secure 40% of global DRAM production capacity, and a $500 million investment in cold treatment research. In response to intensifying competition with Anthropic, it is reported that the company is considering lowering subscription prices. There are also moves to apply for an IPO, and the cryptocurrency market is also drawing attention as an investment opportunity. There are also reports that the White House has requested restrictions on the release of GPT-5.6.

## 4. [Google AI trends] A conversational AI feature called "Ask Maps" has been added to Google Maps based on Gemini, and integration of a food ordering feature is also being considered. GM announced that it will equip the Cadillac and Corvette with Gemini AI. Alphabet's market capitalization is approaching $5 trillion, close to Nvidia's, but some analysts say that increasing AI investment is putting pressure on GOOGL stock. There is little reporting of Sundar Pichai's direct statements.

Microsoftは25億ドル・エンジニア6000人規模の「Frontier Co.」を新設し、企業向けAI実装を専業化した。 Meta teased the upcoming release of a new coding model called Musespark, while Zuckerberg himself admitted that the company's AI restructuring goals "haven't materialized" and that agent progress has slowed. DeepSeek関連ではByteDanceが新たなスケーリング則を発見したと報じられ、AIブームの継続性を支える可能性が指摘されている。 Together AIは8億ドルを調達し、オープンソース推論市場が10億ドル規模を突破した。 xAI/Musk関連は具体的な製品ニュースは少なく、株式・個人ネタが中心。

- Sam Altman (OpenAI CEO): MSN reported that he proposed a 5% stake transfer to the Trump administration and urged other companies such as Google and Anthropic to do the same. Forbes長者番付で資産65億ドルへ急上昇。

- Mark Zuckerberg (Meta CEO): Before announcing its financial results, he admitted that Meta's AI restructuring goals were "not realized" and said that the progress of AI agents was slower than expected. - Peter Thiel (a16z investor): Warns that Anthropic could influence the 2028 US presidential election. - Yann LeCun (Meta Chief AI Scientist): Mentioned in connection with an analytical article that researchers and startups' interest is shifting to "world models." Meta社内ではAI熱狂の中で不安と苛立ちが広がっているとの報道も。 - Jensen Huang (NVIDIA CEO): A leather jacket signed by Jensen Huang was put up for charity auction and has become a hot topic (not a real business trend).

- Geoffrey Hinton: This is related to an article that raises concerns about AI being introduced "too quickly" at a clinic in Yukon Territory, Canada, but this is not a direct statement from the person himself.

Florida's new AI data center law goes into effect, sparking bipartisan concerns. Five Eyes (an information sharing alliance between the UK, US, Australia, Canada, and New Zealand) has expressed the view that ``AI is the best way to combat AI.''カリフォルニア州はAI関連の雇用喪失を追跡する新ツールを導入。消費者法の観点からAI主導の意思決定に関するリスクも指摘されている。

## 9. [Integrated analysis: What is happening now] The axes of competition between OpenAI and Anthropic are shifting to ``distance from politics'' and ``in-house production of hardware.'' OpenAIのトランプ政権への株式提供提案は、規制環境を味方につける戦略と読める一方、Anthropicは逆に政治への影響力行使を懸念される立場に置かれている（Thielの警告）。 Both companies are seeking to break away from dependence on NVIDIA through partnerships with Samsung and their own chips, and this, combined with Alphabet's rapidly increasing market capitalization, is a move that suggests a shake-up of the NVIDIA monopolization system.一方でMetaは1350億ドル規模のAI投資を続けながらも、Zuckerberg自身が成果の遅れを認めるなど、投資と実績のギャップが表面化している。 The new scaling rules from DeepSeek/ByteDance and the funding of Together AI indicate that competition is progressing in parallel on a different axis: open source and low-cost inference. Overall, there is a growing bipolar structure between the three major US companies (OpenAI, Anthropic, and Microsoft), which are highly capital-intensive, and the Chinese and open source camp, which uses cost efficiency as its weapon.

Future points of interest

Timing of realization of custom chip plans for both Anthropic and OpenAI (including partnership with Samsung) 3. Explanation of AI return on investment in Meta financial results announcement and follow-up to Zuckerberg’s remarks