AI Daily News — 2026-07-18 (Sat)

[Overview] The focus of the past 12 hours has been on guarding against Chinese models and securing computing resources for Anthropic. Moonshot AI's Kimi K3 is said to be comparable to OpenAI and Anthropic, and the DeepSeek shock appears to be returning, causing the Nasdaq and S&P 500 to fall by 1%. Anthropic is concurrently negotiating with Meta and SpaceX for computing resources.

[Anthropic/Claude trends] Meta is in talks to lease computing resources to Anthropic for up to $10 billion, NYT and CNBC reported. Computing procurement negotiations with SpaceX have also been reported separately, and it appears that Anthropic's strategy is to prioritize external procurement over expansion of its own data center. While Claude is being introduced to teachers in Detroit, there are reports of backlash against the advertising campaign. It was also confirmed that employees have donated $3 million to support AI safety regulations.

[Trends in OpenAI] In Alabama, two lawsuits have been filed by family members alleging that ChatGPT contributed to the death of a user (the case of a Trafford woman). Mashable reports that Apple has sent legal notices to those who left OpenAI. OpenAI proposes new metrics for evaluating investments and also proposes social safety net policies.

[Google AI trends] Gemini 3.5 Pro announced at Google I/O is still unreleased, and the LA Times reports on the inside story of delays due to coding performance issues and conflicts between teams. On the other hand, Honda and Acura will expand the installation of Gemini to their 2023-2026 models. DeepMind's Demis Hassabis calls for "urgent action" in AI safety standards.

[Trends of other AI companies] Microsoft's Satya Nadella reportedly criticized Anthropic's tendency to reject Claude/Fable in an internal meeting. While xAI is suing users over Grok's creation of deepfake CSAM, it is also facing similar lawsuits. There was an incident where xAI's coding agent exposed the entire repository, and Musk promised "full transparency." 36Kr points out that 840,000 lines of Grok code were open sourced, but traces of the user codebase remained.

[AI celebrities' statements and trends] Sam Altman (OpenAI CEO) has reignited a public feud with Elon Musk and is also facing Apple's departure crisis. Jensen Huang (NVIDIA CEO) denied reports of a delay in the Vera Rubin AI system. Demis Hassabis (Google DeepMind CEO) called for urgent action on AI safety governance, and China's Xi Jinping responded by announcing an AI alliance. Geoffrey Hinton (AI Godfather) gave a talk on AI exceeding human intelligence.

[Breakthroughs and innovations in AI technology] China's Moonshot AI's Kimi K3 has been the biggest topic of discussion, with the BBC reporting on it claiming performance comparable to OpenAI and Anthropic, and the NYT evaluating it as ``threatening America's lead.'' The market reacted immediately, with the Nasdaq/S&P 500 down 1%. The tone that the DeepSeek shock is coming again has been shared in multiple media outlets, including Fortune, Bloomberg, and Business Insider.

[AI regulation/policy/social impact] Illinois enacts AI Safety and Transparency Act for Frontier AI Models. In the U.S. House of Representatives, Rep. Miller-Meeks has introduced a bill (modeled after the Iowa law) that would protect children from predatory AI. The White House has launched "GOLD EAGLE", an AI cybersecurity information sharing hub.

[Integrated analysis: What is happening now] It is clear that the improved performance of Chinese models (Kimi K3, DeepSeek family) has reignited doubts about the computational resource superiority of Western AI companies, putting short-term adjustment pressure on the stock market. Anthropic, on the other hand, is solidifying its defense in the form of computing resource negotiations with Meta and SpaceX, suggesting that it prefers procurement through partnerships rather than expanding its own infrastructure. Microsoft's criticism of Anthropic by Nadella shows that the rivalry between the Copilot and Claude camps has surfaced within the company. On the other hand, OpenAI faces another headwind: product litigation risk (ChatGPT-related death lawsuits), and technological competition and legal and ethical risks are progressing simultaneously. In terms of regulations, Illinois has taken the lead in enacting state legislation, and at the federal level, bills centering on child protection are accumulating.

[Future points of interest] 1. Will an official agreement be reached regarding the computing resource contract between Anthropic and Meta/SpaceX, the scale and conditions? 2. Gemini 3.5 Pro official release date and actual evaluation of coding performance 3. Determine whether the performance verification of Chinese models including Kimi K3 is supported by Western benchmarks, and whether the market reaction is temporary or structural.

AI industry trend report (last 12 hours)

## 1. [Overview] The performance of China's Kimi K3 (Moonshot AI) narrowed the gap with Google Gemini and Fable models, causing market turmoil as the Nasdaq and S&P 500 fell by 1%. It is reported that the US administration is considering a direction in which the government will control access privileges to the top models of OpenAI and Anthropic, and the litigation battle between Apple and OpenAI is intensifying. Meta, Google, and Microsoft continue their arms race with infrastructure investment and talent poaching.

## 2. [Anthropic / Claude trends] Startup Fortune reports that the White House is asserting its position in deciding who gets early access to Claude and GPT's powerful model. An intruder appeared at Anthropic's headquarters and left a warning that ``this person will be killed,'' highlighting the growing threat to AI companies. There are also reports that Meta, led by Zuckerberg, has presented an acquisition offer to Dario Amodei (Anthropic CEO), indicating pressure to poach and integrate the company behind the scenes.

## 3. [OpenAI trends] OpenAI is pursuing a privacy lawsuit seeking dismissal. The conflict between Apple and OpenAI over talent poaching and theft of confidential information has become more serious, and Apple is contacting former employees and taking legal action. Nvidia CEO Jensen Huang said the company's $30 billion OpenAI investment "may be the last before IPO," increasing speculation that the company will go public. Some analysis suggests that the Atlas browser has gone backwards in terms of policy, and this appears to be part of its security measures.

## 4. [Google AI trends] Startup Fortune reported that Alphabet's market capitalization decreased by about $200 billion in one day due to the delayed release of Gemini 3.5 Pro. This is against the backdrop of China's rapid pursuit of AI. NotebookLM announces renaming to "Gemini Notebook" and addition of code tools. There is also a headline saying Google Kills 3 Free Gemini Coding Tools, and the free tier is shrinking. DeepMind CEO Demis Hassabis said that STEM skills make AI tools 10 times more effective.

## 5. [Trends of other AI companies] (Meta, Microsoft, xAI, DeepSeek, etc.) xAI sues users who used Grok to generate child sexual abuse content. It was also discovered that the user's codebase had been uploaded without permission when the Grok build was open sourced, leading to criticism. Information about Grok 4.6/2 trillion parameter class successor model has also surfaced. Reuters reports that DeepSeek is preparing to raise new funds at a valuation of $7.4 billion, a move with an eye on an onshore IPO. Microsoft has announced plans to abolish Outlook's Meeting Insights function and migrate to Copilot, and the trend towards Copilot subscriptions is accelerating. Meta began volume production of its in-house Iris AI chip in September and hired Dave Brown from Amazon to lead AI infrastructure.

## 6. [Remarks and trends of AI celebrities] - Sam Altman (OpenAI CEO): Named as a central figure in the Apple vs. OpenAI human resources and confidential information lawsuit, Sam Altman was reported to be a symbol of security enhancement in response to increasing threats to AI company executives. - Elon Musk(xAI Founder): xAI sues user who misuses Grok to generate child sexual abuse content. SpaceX is reportedly actively helping the Department of Defense develop "lethal AI models." - Jensen Huang (NVIDIA CEO): Expanding AI business development in Japan, stating that ``AI factory is an intelligent production plant that goes beyond a data center.'' Signals $30 billion OpenAI investment as last major investment before going public. - Mark Zuckerberg (Meta CEO): Leads the expansion of AI infrastructure by planning mass production of Iris AI chips and recruiting executives from Amazon. On the other hand, it was also reported that there was an acquisition offer for Dario Amodei. - Demis Hassabis (Google DeepMind CEO): Said that STEM skills improve the effectiveness of using AI tools by 10 times, and emphasized the importance of specialized education in the AI era. - Ilya Sutskever (SSI Co-Founder): Mentioned in the power structure theory of ``Who is really creating AI?'' and brought up in the context of insider dominance theory. - Geoffrey Hinton (AI Godfather): A special article was published about his fear of the AI he created, and he reiterated his concerns about AI safety.

## 7. [Breakthroughs and innovations in AI technology] Bloomberg reports that China's Moonshot AI's new model has narrowed the performance gap with the US model, and the Kimi K3 is also attracting attention. Both models appear to be based on improvements to the Transformer architecture, but detailed technical specifications have not been disclosed. The market took this as a ``breakthrough in Chinese AI,'' and the Nasdaq and S&P 500 fell, indicating that a structure in which technological progress has a direct impact on the stock market is becoming established.

## 8. [AI Regulation/Policy/Social Impact] Colorado repealed its AI anti-discrimination law and replaced it with transparency requirements, but it has been pointed out that the federal government may object. The US administration is reportedly considering placing access to OpenAI and Anthropic's top models under government control, further strengthening national control over AI sovereignty. Concerns about the misuse of AI and the emergence of a surveillance society are also emerging, such as arrests in Florida for producing child sexual abuse content and the approval of large-scale surveillance on highways in North Carolina.

## 9. [Integrated analysis: What is happening now] The rapid performance improvement of Chinese AI (Moonshot AI, Kimi K3, DeepSeek) has entered the stage of directly moving the US stock market. The $200 billion drop in Alphabet stock due to Gemini 3.5 Pro's delay is a typical example, and a structure in which the race to develop AI is directly linked to investor sentiment is becoming entrenched. On the other hand, the US administration is showing moves to place OpenAI/Anthropic model access under national control, and from a geopolitical perspective, the dual pressures of ``strengthening national control of AI sovereignty'' and ``China's catching up'' are progressing at the same time.

Among companies, competition to attract human resources (Meta → Amazon, Apple vs. OpenAI) and infrastructure investment competition (Meta Iris chip, Nvidia deployment in Japan) are parallel, and securing technological superiority is being fought on three fronts: infrastructure, human resources, and capital. Reports of Anthropic's takeover offer for Dario Amodei suggest that pressure from Big Tech to capture leading AI companies is increasing behind the scenes. At the same time, threats to AI company executives (Anthropic headquarters intrusion incident) and crimes abusing AI (Grok CSAM incident) are occurring frequently, and the rapid growth of the AI industry is exposing security and ethical risks.

## 10. [Future points of interest] - The official release date of Gemini 3.5 Pro and the impact of performance comparison with Chinese models (Kimi K3, new DeepSeek model) on market evaluation - Specification of the U.S. administration's access control policy to OpenAI/Anthropic models and the scope of its impact on allied countries and companies - DeepSeek raises capital at $7.4 billion valuation and progresses onshore IPO, expanding capital market access for Chinese AI companies