AI Daily News — 2026-09-23 (Wed) Evening News

AI industry trend report (last 12 hours)

Overview

The announcement of Anthropic's ``Claude Opus 5.5'' and the price reduction of OpenAI's ``GPT-6 Sol/Luna'' coincided at the same time, intensifying price competition. Sam Altman and Dario Amodei are scheduled to address the United Nations Security Council on AI safety, while Trump rejects a global AI regulatory framework and calls it "Super Intelligence," sharpening international divisions over regulation. In China, related stocks have plummeted due to investigative reporting on DeepSeek and Moonshot, and geopolitical risks have also surfaced.

Anthropic / Claude trends

The top model "Claude Opus 5.5" has been officially announced and is attracting attention in terms of benchmarks, price, and safety. In the medical field, the company announced the global expansion of medical AI through a partnership with OpenEvidence (Reuters). It is also gaining recognition in terms of research, with top scientist Jared Kaplan being featured as the person who foresaw the core of the "AI Doom Theory." An IPO observation article (Barron's) has also been published, and investors are increasingly interested in acquiring private stocks.

OpenAI trends

The company is rolling out an aggressive pricing strategy that has successively introduced new models such as "GPT-6 Sol," "GPT-6 Astra," and "Luna," and cut API prices in half. Parallel reported that implementing GPT-6 Astra cut research time and costs in half. On the other hand, organizational issues have also emerged, such as the search for a successor for the chief communications officer (CCO), which has been stalled for nine months. Concerns (NPR) have been reported that an "AI freeze" will benefit large companies and harm small and medium-sized enterprises, and there are also reports of a desire for a single safety standard for the entire industry.

Google AI trends

A Comscore survey reports that ChatGPT's LLM market share is shrinking and Gemini is gaining momentum. There are also prominent analytical articles stating that the battle for AI agent supremacy is intensifying due to the trifecta of Meta Muse, Gemini, and ChatGPT. The Google Cloud/AI business is reported to have grown by 52% year-on-year, and the search business continues to maintain its dominance. Hardware development continues, with the announcement of the Gemini-powered Googlebook Premium notebook PC.

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

Meta is expanding its presence, with "Muse" taking the top spot in the app store and Llama accounting for 38% of modified LLMs. While Microsoft is promoting the effectiveness of 365 Copilot in reducing usage time, it has also received disclosure requests from investors regarding the effectiveness of AI and Copilot (Law360), and it has also been reported that Copilot discounts will be expanded and the "AI Super App" concept. xAI's Grok has surpassed 418K weekly users and is progressing with Tesla integration, but some comparison articles say it is 7 points inferior in performance to Claude. While DeepSeek is focusing on training an 8 trillion parameter model using Huawei chips, Chinese AI stocks have fallen due to investigative reporting on DeepSeek and Moonshot.

Remarks and trends of AI celebrities

- Sam Altman (OpenAI CEO): Will be briefing the UN Security Council on AI safety next week. Today I will be speaking at the United Nations with Dario Amodei. - Dario Amodei (Anthropic CEO): In addition to his UN speech, he also wrote an article on FT arguing that ``AI companies need the same level of supervision as the banking industry.'' It was also brought up in the context of the debate surrounding the AI Doom scenario. - Elon Musk (xAI Founder/Tesla CEO): Emphasized his full investment stance in Nvidia. He predicts that AI will surpass humans in every field within the next two years. - Jensen Huang (NVIDIA CEO): Scheduled to be a special guest at the AI Month event in the capital of Kazakhstan. Established a new medical and scientific AI research center in Israel. He was criticized by Geoffrey Hinton for underestimating the theory of AI slowdown (with the sarcasm that he "has Trump's ears"). - Geoffrey Hinton (AI Godfather/Nobel Prize Winner): Warns that "AI could surpass human intelligence within 5 to 10 years" and refutes Jensen Huang's deceleration theory. - Sundar Pichai (Google/Alphabet CEO): Included in the list to attend the Trump-Xi state dinner. - Mark Zuckerberg (Meta CEO): Plans to announce new AI-era products at Meta Connect conference. - Demis Hassabis (Google DeepMind CEO): Mentioned in the context that "companies are unlikely to pause AI development." There are also reports that a young researcher from DeepMind has transferred to OpenAI. - Ilya Sutskever (SSI co-founder): Only mentioned in the context of AI geopolitics surrounding the United Nations, Trump, and Xi.

Breakthroughs and innovations in AI technology

Cooperative operation of multiple AI agents continues to be a hot topic, with explanatory articles summarizing the architecture and usage examples of "Agent Swarm (multi-agent system)." Google's Gemma 4 12B was introduced as featuring an encoder-free design. Cathie Wood offers an optimistic outlook: ``AI could push real GDP growth up to 7%.'' On the other hand, there are also articles warning that ``slow introduction of AI will lead to lost opportunities'', citing the British Industrial Revolution.

AI Regulation/Policy/Social Impact

New York state is considering AI regulations that include emergency shutdown mechanisms (kill switches). The Trump administration is attempting to change the name from "AI" to "Super Intelligence," and whether or not this will stick is a topic of debate. In the U.S. Senate, Representative Kelly announced the "Make AI Work for Americans Act." The UK announced its commitment to combating disinformation and developing global AI standards. Survey results show that at the board level, a ``lack of AI skills'' is recognized as an issue in most companies.

Integrated analysis: What is happening now

The biggest picture is that ``intensifying price competition'' and ``international division over regulations'' are progressing simultaneously. Anthropic Opus 5.5, OpenAI GPT-6 Sol/Luna, and xAI Grok 4.7 introduced low-cost, high-performance models around the same time, and the decline in inference costs is reshaping the competitive structure of the entire industry (Axios' point that ``cost reduction will help safely slow down'' is suggestive). On the other hand, the Trump administration's refusal to regulate global AI and the change in its name to "Super Intelligence" creates a contrast with Sam Altman and Dario Amodei's speech to the United Nations, in which "major AI companies appeal for safety." On the geopolitical front, there are signs that the decoupling of the US-China AI supply chain will accelerate due to investigative reporting on DeepSeek and Moonshot and deepening dependence on Huawei chips. At the celebrity level, the ``deceleration debate'' of Hinton vs. Huang is symbolic, and the evaluation of the speed of AI progress itself is divided even within the industry. Google Gemini's share is eroding ChatGPT's share, and the composition of the "big four + emerging powers" including Meta Muse and xAI Grok is becoming established.

Future points of interest

The specific content of Sam Altman and Dario Amodei's speech to the UN Security Council and the outcome of the friction between the Trump administration's stance of refusing regulations. 2. Progress in the investigation by Chinese authorities into DeepSeek and Moonshot and the impact of dependence on Huawei chips on China's AI industry. 3. The outcome of the battle for AI agent market share among the four companies Google Gemini, Meta Muse, OpenAI, and xAI Grok, and the impact of competition to lower prices on each company's profit structure.