AI Daily News — 2026-07-08 (Wed)

[Overview] There are signs that the dynamics of the AI industry are changing as Microsoft moves to reduce its dependence on OpenAI and Anthropic by switching to its own model. Geopolitical and supply chain risks have shaken the market, with Nvidia stock falling on reports of DeepSeek developing its own AI chip. Meta has entered the image generation AI competition in earnest with the introduction of Muse Image.

[Anthropic/Claude trends] Bloomberg reports that Microsoft will replace some of the OpenAI/Anthropic models within Copilot with its own AI, and cost reduction appears to be the main reason. On the other hand, Alibaba has banned Claude Code due to security concerns, and it is increasingly being shut out in China. Adoption continues in the government and enterprise fields, including office expansion in New York and the US Cyber Defense Agency (CISA) using Anthropic's AI to scan government software.

[Trends in OpenAI] Legal risks have become apparent, with Canadian provincial governments suing OpenAI over ChatGPT-related gunshot warnings. On the other hand, there have been reports of Australian Payments Plus improving development speed with ChatGPT/Codex and its approach to the medical field (doctors, patients, hospitals), and corporate adoption is steadily expanding. In Knox County, Georgia, there was a meeting between Governor Kemp, OpenAI, and Georgia Power, and collaboration with electric power infrastructure was also a focus.

[Google AI trends] Gemini API is being used to expand the "Managed Agents" functionality, including support for background tasks and remote MCP, and is strengthening the agent platform. Both Cognizant and Accenture are expanding corporate implementation of Gemini Enterprise via Google Cloud, and it was reported that it has also penetrated into the mid-market segment. Analysis suggests that Gemini is in the process of evolving from a simple chatbot to a multi-tool AI platform.

[Trends of other AI companies] (Meta, Microsoft, xAI, DeepSeek, etc.) Meta has launched a new image generation model, ``Muse Image,'' and aims to capture advertisers and subscribers as the first image model after Zuckerberg's AI overhaul. Microsoft is switching to its own model to reduce Copilot costs, and the stock price is rising. xAI has serious problems and deepfake CSAM lawsuits are expanding. DeepSeek is pressuring Nvidia stock with reports of its own AI chip development, attracting attention as a move to avoid US sanctions.

[AI celebrities' statements and trends] - Sam Altman (OpenAI CEO): OpenAI case with Elon Musk has been sent to arbitrator. Trump says he will speed up permitting for fossil fuel power plants for Big Tech. - Dario Amodei (Anthropic CEO): Futurism reports that Anthropic was "secretly monitoring users" and also analyzes Claude's "internal neural patterns." - Jensen Huang (NVIDIA CEO): Cramer mentioned that he described Marvell as "the next trillion dollar company." Bullish predictions regarding Nvidia's growth potential continue. - Sundar Pichai (Google/Alphabet CEO): Jason Calacanis thanks Trump for supporting high-skilled immigration. The United Nations Secretary-General called for a global ban on "killer robots." - Mark Zuckerberg (Meta CEO): He acknowledged that the progress of the Meta AI agent has been slower than expected, and some believed that this suggested the possibility of over-investment in AI. - Demis Hassabis (Google DeepMind CEO): Mentioned in connection with the movement among world leaders to explore the use of AI for economic growth. - Ilya Sutskever (SSI Co-Founder): He expressed the view that achieving human-level intelligence requires more than an LLM. - Yann LeCun (Meta Chief AI Scientist): There were reports related to the founding of a European humanoid robot startup by a former Tesla researcher and the conflict between AI leaders over xAI and market risk. - Geoffrey Hinton (AI Godfather): Mentioned in connection with ``consciousness'' reports that AI has ``constructed its own thinking workspace.''

[Breakthroughs and innovations in AI technology] DeepSeek's new AI acceleration method ("Speed Hack") has become a hot topic, and its improvement in inference efficiency is attracting attention. Meanwhile, a study by Purdue University revealed that U.S. farmers are rejecting the AI revolution, highlighting disparities in adoption by industry. The company is also gaining recognition in the healthcare field, with BostonGene receiving the annual analytics company award at the AI Breakthrough Awards.

[AI regulation/policy/social impact] Illinois Governor Pritzker signed a state law (SB 315) that holds AI companies accountable for safety risks, establishing advanced AI oversight standards in the United States. Norm AI, an AI-related legal startup, has raised $120 million in Series C funding, reaching a unicorn-level valuation of $1.2 billion. Tightening regulations and expansion of the legal tech market are progressing simultaneously.

[Integrated analysis: What is happening now] Microsoft's strategy of prioritizing its own AI means reducing dependence on existing partners such as OpenAI and Anthropic, and signals a turning point when the cloud giant moves to in-house the underlying model. At the same time, Alibaba's ban on Claude Code shows that the AI security conflict between the US and China has extended to the exclusion of practical products. DeepSeek's report on its own chip development has shaken up Nvidia's stock price, and a structure in which the US-China decoupling in the AI semiconductor supply chain is directly linked to investor sentiment is becoming established. On the corporate side, there are signs of a correction to excessive expectations, with Meta and Zuckerberg admitting delays in the progress of AI agents, while Google and Microsoft are steadily building up their agent base and enterprise deployment. On the regulatory front, the new Illinois law has become a new reference point for AI safety regulation in the United States, and along with the rapid growth of the legal AI market (Norm AI), there is a growing trend for AI governance to become a business.

[Future points of interest] 1. How will Microsoft's transition to its own AI affect its contracts and revenue structure with OpenAI and Anthropic? 2. Progress of DeepSeek's proprietary AI chip development and impact on Nvidia/US export regulations 3. Will regulations spread to other states starting with the Illinois AI Safety Act?

AI industry trend report (last 12 hours)

## 1. [Overview] Security controversy escalates to a geopolitical level after Anthropic receives 'backdoor' warning from Chinese authorities regarding Claude Code. At the same time, OpenAI reportedly received approval from the US administration for the widespread rollout of GPT-5.6, and xAI announced that Grok 4.5 will be released to the public on July 9th. There are also reports that Microsoft is considering switching to its own AI, and the battle for leadership among major companies is intensifying.

## 2. [Anthropic / Claude trends] CNBC, Reuters, and WSJ all reported that China discovered a "backdoor" security vulnerability in Claude Code. Chinese authorities have issued warnings about AI risks. Meanwhile, Anthropic has removed a hidden Claude Code tracker after researchers pointed it out, Decrypt reports. Claude Cowork (a feature that allows you to remotely control AI work from your smartphone) was also newly introduced. A $19 billion deal with TeraWulf is also in the works.

## 3. [OpenAI trends] Axios reports that the Trump administration has lifted restrictions on GPT-5.6, and Reuters also reports that widespread rollout is imminent. Meanwhile, OpenAI's Chief Futurist, Joshua Achiam, left the company after nine years. The South China Morning Post reported that a Canadian province is preparing to sue OpenAI in the wake of a school shooting, raising the issue of legal liability for AI companies once again.

## 4. [Google AI trends] Gemini 3.5 Pro is reportedly scheduled to be released on July 17th, in contrast to DeepSeek's July 24th deadline. Gemini AI privacy class action lawsuit dismissed by US federal judge. Cognizant expands partnership with Google Cloud to accelerate enterprise adoption of Gemini Enterprise. Google Marketplace has opened a new channel for publishing AI agents.

## 5. [Trends of other AI companies] (Meta, Microsoft, xAI, DeepSeek, etc.) Meta launches image generation model "Muse Image" by integrating it with Instagram and WhatsApp. As announced by Musk, xAI plans to release Grok 4.5 on July 9th, claiming "Opus-class" performance. SiliconANGLE and 36Kr report that Microsoft is considering switching from OpenAI and Anthropic models to in-house AI (cost reduction is said to be the reason). DeepSeek is developing its own AI inference chip and is attracting attention as a potential rival to Nvidia.

## 6. [Remarks and trends of AI celebrities] - Sam Altman (OpenAI CEO): Rebutting skeptics who call LLM a "dead end." Ahead of the launch of the science-specific GPT-5.6, it has been reported that the US government wants to own a stake in OpenAI. - Elon Musk (xAI Founder / Tesla CEO): Announced public availability on July 9th, calling Grok 4.5 an "Opus-class" model. Meanwhile, Tesla's stock price has fallen following Musk's $25 billion AI investment, causing investor concern. - Jensen Huang (NVIDIA CEO): Signed $27 billion partnership with chip competitor as part of "AI Factory" strategy. This was also a positive sign for Micron's stock price. - Dario Amodei (Anthropic CEO): Claude Code's hidden tracker removal has become a hot topic amid renewed reports of opposition to the Trump administration's AI licensing policy. - Sundar Pichai (Google/Alphabet CEO): Hundreds of students walk out in protest of Stanford University's commencement speech. - Mark Zuckerberg (Meta CEO): While there are reports of tests of "super sensing" AI glasses, Chamath Palihapitiya criticizes Meta's AI strategy as a "total failure". - Yann LeCun (Meta Chief AI Scientist): Expressed the view that current AI is "not true intelligence." French startup ZML releases a free inference tool with his support.

## 7. [Breakthroughs and innovations in AI technology] DeepSeek's decision to develop its own AI inference chip is attracting attention as a direct challenge to Nvidia. SambaNova has reached a valuation of $11 billion as a rival to Nvidia chips. ZML (a French startup backed by LeCun) has released a tool for accelerating inference between multi-vendor AI chips for free, intensifying competition in the infrastructure layer.

## 8. [AI Regulation/Policy/Social Impact] Illinois enacts the strictest AI safety law in the nation. On the other hand, Forbes points out that the proliferation of AI regulations by state is forcing companies to deal with compliance in each jurisdiction. In Canada, preparations are underway to file a lawsuit against OpenAI in response to the school shooting incident, and Bloomberg Law is also raising questions about the scope of agent-type AI's legal liability.

## 9. [Integrated analysis: What is happening now] The biggest structural change is Microsoft's "in-house AI shift" report. If true, this could be seen as a move to reduce dependence on OpenAI/Anthropic and lower costs, and a sign that major cloud vendors are becoming independent from underlying model providers. At the same time, the Chinese authorities' warning of a "backdoor" to Claude Code shows that the AI security friction between the US and China has escalated into mutual distrust at the product level. The release of Grok 4.5 and GPT-5.6 around the same time creates a direct confrontation between xAI and OpenAI, and Musk is also drawing Anthropic into the center of the competition by calling itself Opus-class. DeepSeek's in-house chip development suggests that China's AI independence strategy is progressing not only in the model layer but also in the hardware layer. On the investment front, SambaNova's rapidly increasing valuation and Nvidia's investment strategy in chip companies reflect the shift towards a GPU-dominated structure. Regulations are becoming increasingly fragmented on a state-by-state basis, and companies are becoming entrenched in having to bear the costs of responding to each jurisdiction.

## 10. [Future points of interest] 1. Performance comparison competition between Grok 4.5 (July 9th), Gemini 3.5 Pro (targeted July 17th), and DeepSeek new model (deadline July 24th) 2. How far will Microsoft's move away from relying on OpenAI/Anthropic models actually progress with Copilot products? 3. How will the Chinese government's Claude Code security warning affect U.S.-China AI trade and regulatory friction?