AI Daily News — 2026-10-01 (Thu) Morning News

AI industry analysis report (last 12 hours)

Overview

The FTC has launched a large-scale investigation into the product risks of "super intelligence" models, targeting companies including Anthropic and OpenAI. At the same time, voluntary AI standards were agreed upon at a "super intelligence" summit between President Trump and major AI executives. However, doubts about the binding force and uncertainty about safety remain. DeepSeek and Huawei's software partnership for AI chips is also attracting attention as a geopolitical challenge to Nvidia's dominance.

Anthropic / Claude trends

- FTC investigates Anthropic and OpenAI. Multiple major media outlets (Reuters, CNBC, WSJ, NYT, CBS) have reported on the issue, which focuses on the potential harm to consumers and product risks. - Anthropic claimed that a popular Chinese AI model has "Mythos-class" hacking capabilities (Tom's Hardware). The details of the word Mythos are unclear from the headline alone. - It was reported that the company warned of the "existential risk" of AI in its IPO application documents (LA Times). This suggests that Anthropic is in the preparation stage of listing. - Regulatory pressure and IPO preparations are progressing at the same time, and companies are now faced with the challenge of balancing risk disclosure and business growth.

OpenAI trends

- OpenAI announced on its official blog that it had disrupted a "coordinated model-distillation campaign." This content shows a defensive stance against the leakage of intellectual property due to model distillation. - Sam Altman reportedly announced a personal AI agent called "Dots" (Patch, spyglass.org). - Altman expected to skip Congressional hearing on runaway AI agents (NBC). The company is currently under investigation by the FTC, which could lead to criticism of its response. - There are reports that the company has considered raising funds from China and Russia in the past, as well as considering a GPT-3 preview for Kanye West. This is said to be the background to Amodei's withdrawal, and the details have not been confirmed. - AI utilization support for small businesses is also announced on the official blog.

Google AI trends

- Introducing reusable "skills" to Gemini and replacing Gems. The policy is not to erect a paywall (Unite.AI, PCWorld). - Extended GSA's OneGov contract for Gemini, solidifying its position in the government market (FedScoop). - Expand Gemini usage in Asia through partnership with Antom. Android is moving from Google Assistant to Gemini, and user reactions are being reported. - Google DeepMind has released SynthID Bio for AI-generated proteins.

Trends of other AI companies

- Microsoft: Default pay-as-you-go for Copilot Business licenses. "Agents are very sensitive customers," one executive said, warning that many apps would lose pricing power. Some estimates suggest that Copilot could generate $2 billion in annual revenue by 2028. - xAI(SpaceXAI): Considering a four-tier pricing structure for Grok and X users (Bloomberg). - DeepSeek: Jointly develop software for open source AI chips with Huawei. Aiming to reduce Nvidia's dependence on CUDA, it will proceed with optimization for Ascend (NYT and others). - Others: CoreWeave provides NVIDIA Vera CPUs for agents. Cloudflare has announced a tool to charge AI agents with stablecoins. New Meta-specific news is scarce.

Remarks and trends of AI celebrities

- Sam Altman (OpenAI CEO): Announced "Dots". He is expected to be absent from the Congressional hearing. - Dario Amodei (Anthropic CEO): AI has great benefits, but risks need to be addressed (Economic Times). IPO documents refer to existential risk. - Elon Musk (xAI Founder / Tesla CEO): AI will bring wealth and better healthcare, but said it will change the shape of work. It also refers to skilled data center workers. - Jensen Huang (NVIDIA CEO): "If you're concerned, just don't do it," he told AI Lab. It was reported that there was a lot of exposure in Q3, which was reflected in the stock price. - Sundar Pichai (Google CEO): Supported the White House AI Safety Agreement. Meanwhile, there are reports that Trump called him a "monster" at the summit. - Mark Zuckerberg (Meta CEO): Presented AI framework after meeting with Trump. - Satya Nadella (Microsoft CEO): Sat next to Bezos at a luncheon hosted by Trump. - Demis Hassabis (Google DeepMind CEO): Protein watermark research related to DeepMind, such as SynthID Bio, was reported. This is not his own statement. - Geoffrey Hinton (Nobel Laureate): Warned that 10% extinction risk is "not unreasonable." Voluntary safety pledges are ridiculed even though they are said to be "binding." - Ilya Sutskever / Yann LeCun: There are no direct statements from the person himself, only related reports.

Breakthroughs and innovations in AI technology

- SynthID Bio demonstrated a method to add watermarks to AI-generated proteins while preserving their functionality (Nature, DeepMind). This is a step towards provenance management in the bio field. - Chutes AI trained an 8B model on a distributed gaming GPU and ran it on a smartphone's CPU. - A small model was created that uses a single RTX 3080 Ti to "play" Pokémon Red with next-state prediction. This can be said to be an example of a world model approach.

AI Regulation/Policy/Social Impact

- The White House has directed federal agencies to move into the era of "super intelligence" (IAPP). - Trump and major AI leaders agree on voluntary AI standards. It also became a hot topic that "United States" was misspelled in the signed pledge document (TechCrunch). - NYT Op-ed questions AI giants' policy of self-regulation. - The FTC investigation is in full swing, and the focus of regulation is on two fronts: self-regulation and law enforcement. - It has been reported that a new AI regulation law will come into effect on October 1st. Details are unclear from the headline. - There are also concerns that the Global South will be left behind.

Integrated analysis: What is happening now

In terms of policy, there appears to be a dual structure in which the government promotes self-regulation while the FTC pursues law enforcement in parallel. The effectiveness of the leaders' security agreement is being questioned, and the gap between Mr. Hinton's warning and Mr. Huang's optimism is symbolic.

In terms of power, Anthropic's IPO preparations contrast with OpenAI's distillation measures and Dots' consumer rollout. Google is pushing Gemini into the Android and government markets, and Microsoft is moving to a pay-as-you-go, agent-era revenue model.

On the technical side, the issue of price and revenue is moving from ``per seat'' to ``per usage/agent'' (Microsoft, Cloudflare, and Grok's 4-step proposal). The problem of distillation and the Chinese model's claims of competence raise the importance of model defense.

In geopolitics, DeepSeek's partnership with Huawei shakes up Nvidia's software dominance. The divide between the US and China is extending from chips to the underlying software layer.

Please note that the above analysis is based on headlines only, and the text of the article has not been checked. Details are not definitive information.

Future points of interest

Scope of FTC Investigation and Target Company Response, and Impact on Anthropic's IPO Process. 2. Will the performance and popularity of DeepSeek and Huawei's software infrastructure disrupt Nvidia's CUDA advantage? 3. The effectiveness of voluntary AI standards and the future of the debate over runaway agents in Congress.