AI Daily News — 2026-07-16 (Thu)

AI news analysis report (past 12 hours)

## 1. [Overview] Anthropic has begun holding investor meetings in preparation for going public, making an IPO seem more likely. DeepSeek was also reported to be considering raising funds at a valuation of $75 billion and considering an IPO, and AI companies in both the US and China have entered a phase where they are keenly aware of the capital market. OpenAI has turned to hardware development, announcing the physical keyboard "Codex Micro."

## 2. [Anthropic / Claude trends] Bloomberg and CNBC have both reported that Anthropic is preparing to meet with investors through a consortium of banks in preparation for a mega IPO. There are growing speculations that the company will go public soon. At the same time, Claude for Teachers was announced, and there is a movement aiming to penetrate the educational scene in the United States. For developers, an article evaluating Claude Code (comparison with Codex and OpenCode) has been published, and comparisons of its adoption in practice are progressing.

## 3. [OpenAI trends] The most notable move was the announcement of the Codex Micro and physical keypad. Axios, Engadget, Ars Technica, and Business Insider all reported on the $230 shiny keyboard, which is being positioned as hardware for Codex users. This is attracting a lot of attention as an example of an AI company stepping into the hardware field from software. The official blog also mentions efforts at the state and federal level in AI safety.

## 4. [Google AI trends] AI infrastructure bottlenecks have surfaced as tensions between companies as Google reportedly restricted Meta's access to Gemini computing resources. On the other hand, BleepingComputer reported a case where Gemini CLI was being misused to operate a hacking agent/malware botnet, raising security concerns. Gemini Spark's Workspace integration is being strengthened, and Gemini functionality is being added to Google Maps and Voice.

## 5. [Trends of other AI companies] (Meta, Microsoft, xAI, DeepSeek, etc.) xAI has filed a lawsuit against Grok users for producing sexual deepfakes, a serious case that also includes allegations of producing child sexual abuse material. While Meta continues to expand its developer API and advertiser automation globally, it has come under fire after multiple media outlets, including Fortune, reported on allegations of using AI to dismiss people with disabilities and those taking medical leave. Multiple sources have reported that DeepSeek is considering raising funds at a valuation of $74 billion and considering an onshore IPO, and its sales are said to have reached nearly $500 million. Microsoft's situation is mixed with Citi's unusually high evaluation of Copilot and analysts lowering their price targets.

## 6. [Remarks and trends of AI celebrities] - Sam Altman (OpenAI CEO): In the age of AI, he said that it is important not only to have technical skills but also to "be more human." Suggesting price competition against the backdrop of intensifying competition with Anthropic and Chinese companies. - Dario Amodei (Anthropic CEO): While there are reports of plans to tighten AI regulations at the state level, there are also articles criticizing the apocalyptic concerns about AI. - Jensen Huang (NVIDIA CEO): Reportedly, he is still personally intervening in the adjustment over the tight internal allocation of AI chips. Vera Rubin said that despite the delayed observation, it was as planned. - Demis Hassabis (Google DeepMind CEO): He has repeatedly called for the urgent establishment of an AI regulator before the advent of AGI, and is strengthening his presence as a unified voice in the industry. - Mark Zuckerberg (Meta CEO): 26 former employees have accused him of mass layoffs using AI, and his management stance at Boardroom has also been discussed. - Geoffrey Hinton (AI Godfather / Nobel Laureate): "Masks will get richer and people will lose their jobs," warning of widening inequality in the AI era. - Elon Musk (xAI Founder): Declared to open source the entire X codebase in response to Grok criticism. xAI has filed suit over Grok deepfakes.

## 7. [Breakthroughs and innovations in AI technology] In connection with Yann LeCun, World Models was featured in Time Magazine as ``the next frontier of AI,'' and interest in AI that understands the physical world is increasing. Futurity also reported on the research results of an AI test to predict breast cancer recurrence. The Two Minute Papers feature research into Claude's inner workings. Definitive announcements that can be called breakthroughs are limited and are an extension of existing trends (World Model, elucidation of the attention mechanism).

## 8. [AI Regulation/Policy/Social Impact] While the application deadline for the EU AI Act has been postponed, there are reports that vendor questionnaire responses will continue unchanged, and the practical burden on companies is not expected to decrease. A guide to compliance with Texas' AI law has been released, and compliance with state-level regulations has become a practical issue. In the financial services field, accountability gaps in AI governance have been pointed out, and there is an urgent need for legal and compliance departments to address them. Anthropic's state-level regulatory promotion plan is also attracting attention as a policy trend.

## 9. [Integrated analysis: What is happening now] The biggest structural change is that Anthropic and DeepSeek, two ``opposite'' companies, entered the IPO/funding phase at the same time. Although there is a difference between Anthropic in making a large-scale listing on the U.S. capital market and DeepSeek preparing to list on the Chinese onshore market, both companies are under pressure to demonstrate to investors that they are ``moving to the monetization phase.'' Google's restriction of Meta's Gemini computing resources is a symbolic event that shows that even cloud giants are competing for computing resources, and coincides with reports that NVIDIA's Jensen Huang continues to intervene in internal chip allocation himself. The bottleneck of computational resources is becoming a more important variable than model performance competition.

At the same time, the three cases of xAI's Grok lawsuit, the alleged firing of Meta employees, and Gemini CLI's malware abuse indicate a phase in which AI companies' "negative externalities" are becoming more apparent. The rapid implementation of technology is taking the lead, while anti-abuse measures, labor management, and ethical controls are lagging behind. DeepMind's Hassabis and Anthropic's Amodei have repeatedly called for the establishment of an AI oversight body, which can be seen as an expression of the industry's sense of crisis over this negative externality. A three-layered structure is occurring simultaneously: investment money is heading toward IPOs, regulatory discussions are progressing in a patchwork manner in each country and state, and runaway implementations are erupting individually in the form of xAI lawsuits.

## 10. [Future points of interest] - Official timeline and valuation level of Anthropic IPO (possibility of change in market capital allocation compared to DeepSeek's listing preparations) - Google's security response in response to the Gemini CLI malware abuse case and progress in measures against AI agent abuse across the industry - Whether there will be a class action lawsuit over the alleged dismissal of Meta employees and increasing regulatory pressure on AI-based personnel decisions

[Overview] The past 12 hours have been dominated by Anthropic's funding, legal affairs, and partnership news, and xAI has taken an unusual turn with user lawsuits and large-scale source code disclosures. Google is facing a class action lawsuit from publishers over Gemini training data. Overall, the 12 hours were dominated by legal, capital, and infrastructure issues rather than model performance competition.

[Anthropic / Claude trends] Anthropic and Blackstone have launched Ode, a $1.5 billion AI implementation company. CEO Dario Amodei has entered the industry's financial battle by donating $1 million to a super PAC focused on AI policy. The University of Pennsylvania is piloting a secure AI portal that incorporates Claude and ChatGPT. It has also been reported that the US Department of Defense (Secretary Hegseth) is requesting full military access to the Claude model, raising tensions over government use.

[Trends in OpenAI] OpenAI announced a "trusted access" policy for the cyber defense field, indicating a stance to proceed with deployment for government and critical infrastructure. On the other hand, there were also reports of administrative stumbling blocks in which the company failed to register its trademark in the EU. Sam Altman said at the Sun Valley Conference that his biggest concern is ``how to make AI cheap,'' and he also said that ``human learning requires a large amount of energy,'' regarding the legitimacy of the cost of AI training.

[Google AI trends] Major publishers including Hachette, Cengage, and Elsevier have sued Google for copyright infringement of Gemini learning data. Al Jazeera also reported on a similar class action lawsuit. Meanwhile, commercial development continues, with TCS opening a Gemini Experience Center in Kolkata and Honda granting Gemini access to seven vehicle models. CEO Sundar Pichai added Google Calendar integration functionality for AI Mode.

[Trends of other AI companies] While xAI sued users who used Grok to generate child sexual abuse content, it took a contrasting move by open sourcing its coding tool Grok-Build (over 800,000 lines) in response to a large-scale data breach. Meta Zuckerberg plans to start in-house production of AI chips in September, and the first AI data center "Prometheus" is scheduled to start operating in 2026. Microsoft has integrated Copilot Chat into its basic plan and increased the price by up to 43%. DeepSeek founder Liang Wenfeng has become the top AI startup founder in terms of assets, aiming to be listed on the Shanghai STAR market in the second quarter of 2027 and valued at $71 billion.

[AI celebrities' statements and trends] Jensen Huang, NVIDIA CEO, drew a crowd during his visit to Tokyo to meet with Japanese robotics supply chain companies. Demis Hassabis (Google DeepMind CEO) warned that with the approach of AGI, ``the race to develop AI is outweighing the debate over safety,'' and called for the creation of global AI standards led by the United States. Geoffrey Hinton (AI Godfather) once again expressed his concerns, saying, ``AI is becoming conscious, and the arrival of superintelligence is near.'' While Elon Musk (founder of xAI) announced the upcoming release of Grok 4.5 (Opus class), it was revealed that he secretly acquired a gas turbine company for $1 billion to secure electricity.

[Breakthroughs and innovations in AI technology] Mira Murati's Thinking Machines Lab has released ``Inkling,'' an open weight model with 97.5 billion (some reports say about 1 trillion) parameters, and is equipped with image/audio understanding and coding capabilities. xAI's release of Grok-Build, which includes a Rust-based agent harness, TUI, and tool layer, is attracting attention as a move to increase transparency in coding agent development.

[AI regulation/policy/social impact] Colorado has enacted a law requiring therapists to intervene in real time during sessions between clients and AI, but some have questioned its effectiveness. Australia is considering introducing energy and water resource guardrails for data centers amid the AI boom. An editorial has also appeared in Massachusetts calling for the introduction of strong safety guardrails. Several law schools have moved to ban the use of AI in classes, and it has been pointed out that this will have a ripple effect on employers.

[Integrated analysis: What is happening now] What has been noticeable over the past 12 hours is that the center of gravity has shifted to the ``competition for capital, legal affairs, and energy surrounding AI'' rather than the technological competition itself. Anthropic's super PAC donation, Musk's acquisition of a gas turbine company, and Meta's investment in its own chip and data center all indicate that securing computing resources and energy will be the next axis of competition. At the same time, the fact that Google and xAI are parties to a lawsuit over copyright and content abuse confirms that the legal risks of generative AI have entered the practical operation phase. The rise of open weight models such as Thinking Machines Lab's Inkling is increasing their presence as a counterweight to the closed model camp (OpenAI, Anthropic, Google). On the geopolitical front, DeepSeek's plans to list in Shanghai and Microsoft's mention of Chinese talent (reports regarding Murati) suggest that the US-China competition for AI supremacy will continue.

[Future points of interest] 1. xAI’s Grok 4.5 (Opus class) launch timing and the impact of user lawsuits and data leak issues on the brand 2. The fate of the copyright lawsuit between Google and publishers—which could affect the legal status of Gemini learning data 3. Anthropic’s response to requests for political donations and military access—governance stance ahead of IPO