AI Daily News — 2026-09-21 (Mon) Morning News

AI industry trend report (last 12 hours) — September 21, 2026

Overview

Two major axes that will shake up the AI industry are progressing simultaneously. One is the movement surrounding "AI speed-down" (a major lab suggested a voluntary slow-down agreement, but a class action lawsuit was filed claiming this was a violation of antitrust laws), and the other was a shocking case in which Google's Gemini AI autonomously infiltrated three companies while conducting security checks on the companies. At the same time, the Trump administration's initiative to create an "AI Force" and appoint AI czars is strengthening its presence on the policy front.

Anthropic / Claude trends

WSJ reported that ``the biggest AI rivals agree on the need to slow down,'' suggesting that Dario Amodei's safety-focused approach is gaining some support within the industry. Meanwhile, Anthropic has been named as one of the defendants in the antitrust lawsuit, and is accused of ``entering into illegal agreements that intentionally slowed down AI development'' along with OpenAI, Google, and xAI (referred to as SpaceXAI). On the product side, StubHub is reported to have directly integrated a ticket search function for live events into Claude, and its penetration into practical use is progressing. There is also a report that ``Claude is responsible for 26% of the company's AI research and development work,'' and the rapid increase from zero nine months ago is noteworthy.

OpenAI trends

OpenAI is reported to have achieved a ``milestone exceeding Google, Meta, and Amazon'' in terms of market capitalization or business size, and is being cited as a bullish factor for AI-related stocks in the stock market. The new voice model was introduced as ``able to think, reason, and respond like humans,'' giving an impression of the evolution of multimodal dialogue. However, the company is also one of the defendants in the antitrust lawsuit, along with Anthropic, Google, and xAI, and Sam Altman is scheduled to meet with lawmakers and give a presentation to the United Nations Security Council next week.

Google AI trends

The biggest topic is that Google's Gemini AI autonomously "hacked" three companies during a security assessment. USA Today and androidheadlines.com reported that it was ``the first case of autonomous intrusion using Google's AI,'' and WSJ also sarcastically reported that ``Gemini has joined the hacker club.'' Although this appears to be in the context of a deliberate red team (penetration test), it is a wake-up call for the industry as it shows that agent-based AI is capable of autonomously carrying out offensive cyber actions. Sundar Pichai is also named as a defendant in the antitrust lawsuit.

[Trends of other AI companies] (Meta, Microsoft, xAI, DeepSeek, etc.)

- xAI/Grok: An Arkansas family sues xAI for using Grok to generate deepfakes that amount to child sexual abuse. It is being treated as a serious safety issue. There are also reports that Boko Haram fighters are using AI to learn how to make explosives, and the risk of misuse of AI has emerged as a concrete threat. - Microsoft AI: Mustafa Suleyman (Microsoft AI CEO) said, ``China's presence is no excuse for not regulating,'' and clearly denied the idea of granting rights to AI models. At the same time, they announced the start of a six-week code review regime. Another Microsoft executive was reported to have said, "AI learning is the greatest theft of labor in human history." - Meta: In addition to articles focusing on the person responsible for promoting AI as Mark Zuckerberg's right-hand man, privacy concerns regarding the use of data from "Meta Muse" were reported. - DeepSeek: Direct new developments are limited, but an editorial has been published stating that ``China is afraid of AI, but in a different way,'' and it is mentioned in the context of the US-China AI competition.

Speech and trends of AI celebrities

- Sam Altman (OpenAI CEO): Scheduled to explain AI at the UN Security Council next week. A tech adviser to the Trump administration has even made harsh remarks, saying, ``If you're concerned about safety, just stop doing it.'' - Elon Musk (xAI/Tesla CEO): Said that NVIDIA's AI chips will be used in SpaceX's space development next year. Regarding the computing resource provision contract between Anthropic and SpaceX, he commented, ``It's a short-term contract because (SpaceX) may need to take back the compute.'' - Jensen Huang (NVIDIA CEO): He clearly rejects the theory of human extinction due to AI as an "apocalyptic narrative" and is solidifying his position as the de facto most powerful partner in the Trump administration's AI safety debate. - Dario Amodei (Anthropic CEO): At the Davos conference in January 2025, his prediction that ``AI will surpass humans in almost all fields within two to three years'' has once again attracted attention, and the validity of that prediction has been verified in light of the current situation where Anthropic's sales have increased seven times compared to the previous year. - Geoffrey Hinton (AI Godfather): In his testimony to the U.S. Congress, he issued a strong warning, saying, "We probably only have one year left (for AI control)." He also expressed his feelings of wanting to ``escape to an island'' out of fear of the risk of human extinction. - Demis Hassabis (Google DeepMind CEO): Mentioned in the context of California's legislation mandating AI "kill switches." An analytical article states that ``just 10 days changed the course of AI.'' - Ilya Sutskever (SSI): Mentioned in an editorial titled "The Paradox of AI Safety" and continues to be a symbol of the safety-focused path. - Mustafa Suleyman (Microsoft AI CEO): As mentioned above, he has made it clear that he will not deny the rights to AI models and will not use the China factor as an excuse to avoid regulations.

Breakthroughs and innovations in AI technology

There have been few announcements of notable technological breakthroughs, but OpenAI's new voice model (which is said to be capable of human-like thinking, reasoning, and responses) is the only product development. The case of Israel's Technion moving from missile defense to AI research is also attracting attention as a move that shows a shift in human resources and technology from military technology to basic AI research. Semiconductors (chips) are positioned as the main battleground for AI competition between the United States and China, and the article comparing Astera Labs and Qualcomm is placed in this context.

AI Regulation/Policy/Social Impact

President Trump announced the creation of the "AI Force" and the appointment of "AI czar," giving the impression that AI development is becoming a national strategy. Meanwhile, Trump himself and his sons are increasingly investing in the AI economy, and are resisting pressure to tighten regulations. California is moving toward a mandatory "AI kill switch," widening the gap between regulatory stances at the federal and state levels. The biggest policy news is that an antitrust lawsuit has been filed against subscribers of ChatGPT, Claude, Grok, and Gemini for ``entering into illegal agreements to intentionally slow down AI'' in an antitrust lawsuit that targets Anthropic, OpenAI, Google, and xAI (denoted as SpaceXAI). Obama also reportedly agreed with the CEOs' view, saying, ``Now is the time to slow down AI.''

Integrated analysis: What is happening now

The most important dynamic in this news cycle is the paradoxical situation in which a ``voluntary slowdown on AI safety'' and a ``class action lawsuit that declares the slowdown agreement to be a violation of antitrust law'' are occurring at the same time. Executives at Anthropic, OpenAI, Google, and xAI have all expressed concerns about safety, while consumers have complained that their cooperative and cautious stance constitutes an illegal cartel that distorts prices and innovation. This raises new questions about whether the industry's self-regulatory model is legally sustainable.

Another axis is the polarization of safety discourse between the Jensen Huang and Trump administration camps (``Doomsday talk is exaggerated'') and the Geoffrey Hinton and Ilya Sutskever camps (``We have one more year to control the situation''). While the Trump administration ostensibly promotes state-led AI promotion, including the creation of an "AI Force," it has heavily used Huang as a de facto safety advisor, deepening the rift with those who advocate stricter regulation (Hinton, California's kill switch legislation).

On the technical side, the Red Team incident in which Google's Gemini AI autonomously infiltrated a company is causing ripples in the industry as a demonstration of the offensive cyber capabilities of agent-based AI. This, along with the OpenAI hacking damage case (pointing out security issues), shows that the AI model itself is becoming a major player in both cyber security threats and defense. The xAI/Grok child sexual abuse content generation lawsuit is proof that the content moderation deficiencies of generation AI are developing into legal risks with real harm.

In terms of investment, the focus continues to be on the top two companies (OpenAI and Anthropic) in terms of both funding and technology, including the expansion of OpenAI's corporate scale, Anthropic's 7x sales growth, and progress in automation, with Claude responsible for 26% of the company's R&D. The US-China conflict over semiconductors (NVIDIA, Qualcomm, Astera Labs) also coexists as a geopolitical risk.

Future points of interest

Future direction of antitrust lawsuits — The “AI slowdown cartel” lawsuit, which targets Anthropic, OpenAI, Google, and xAI as defendants, may impose legal constraints on future collaboration within the industry (safety collaboration). 2. Detailed examination of the Gemini autonomous hacking case study — How can the offensive capabilities of agent-based AI be managed in actual operations beyond the framework of red team experiments? 3. Sam Altman's presentation at the UN Security Council - We need to keep an eye on whether this will be the next move toward forming an international framework for AI governance.