AI Daily News — 2026-06-25 (Thu)

AI Industry Trend Report (June 25, 2026)

Overview

Three of the biggest topics in the AI industry over the past 12 hours. OpenAI announced its proprietary inference chip "Jalapeño" jointly developed with Broadcom, accelerating its break away from dependence on NVIDIA, and Anthropic accused Alibaba of unauthorized access to its models, creating a new phase in the US-China AI competition. Additionally, Grok has come under fire for its adult content-driven traffic, while the ethics debate surrounding military AI is heating up after Elon Musk's Grok was reportedly used to guide missiles to attack Iran.

Anthropic / Claude trends

Most important: Alleged unauthorized access to AlibaBa Anthropic accuses Alibaba of "unauthorizedly" extracting Claude AI's modeling capabilities (Bloomberg, Reuters). Specifically, it is said that Claude's knowledge was obtained illegally through a method that appears to be model distillation, and the issue of AI technology leakage between the US and China has a significant political impact.

Other trends - Anthropic retiree establishes new startup specializing in AI development support for scientists (WSJ) - The Atlantic publishes an article asking Claude if he would ``refuse illegal military orders,'' drawing attention to the ethical boundaries of AI. - A man living in California reports damage caused by unauthorized use of Claude API account, security issue comes to light

OpenAI trends

Announcement of original AI chip “Jalapeño” OpenAI and Broadcom jointly announced a proprietary chip optimized for LLM inference. CNN reports that ChatGPT's performance improvements are part of a "full stack build" strategy, while TechCrunch reports. The aim is to reduce dependence on NVIDIA GPUs, and Sam Altman has taken a clear stance against NVIDIA.

G7 Summit Attendance Sam Altman, Dario Amodei, and others all participated in President Trump's G7 summit, demonstrating their policy presence (Stocktwits).

Google AI trends

Introducing Computer Use feature in Gemini 3.5 Flash Google DeepMind announces Computer Use feature for Gemini 3.5 Flash on its official blog. The ability for AI to directly control a PC has been implemented in the Gemini series.

Gemini 3.5 Pro delayed to July The launch of the next-generation frontier model Gemini 3.5 Pro has been postponed to July due to adjustments (Business Insider).

others - Start selling AI home speakers equipped with Gemini - Nokia integrates Google Cloud's Gemini agent into network operations - Sundar Pichai tells students worried about job security in the age of AI to "choose optimism" at Stanford graduation ceremony

Trends of other AI companies

Meta - Mark Zuckerberg confirms multi-generation processor deal with Qualcomm - AI-powered prediction market app development plan leaked from internal documents (VPM) - Employee tracking tool suspended due to privacy concerns for AI learning use (The Guardian) - Meta and Microsoft commit to tens of billions of dollars in data center leasing investments

xAI/Grok - Most of Grok's traffic comes from adult content (Forbes, The Information) - Grok Build implements MongoDB plugin and autonomous execution functionality (Plan → Execute → Verify) - Most important: Report that the Pentagon used Grok AI to launch 2,000 missiles into Iran (AOL). The actual combat operation of military AI has become a reality.

Microsoft - Microsoft Copilot helps uncover cybercrime tools (Bloomberg) - Announcement of water usage reduction results in the AI era (Axios)

DeepSeek - Tencent subsidiary WeCom deploys DeepSeek-powered AI agent, stock price rises 6% - Anthropic's model distillation issue has reignited the debate that export controls alone won't stop China from following suit.

Remarks and trends of AI celebrities

| People | Trends | ｜ーー｜ーー｜ | Sam Altman (OpenAI CEO) | Clarified independence from NVIDIA with Jalapeño chip announcement. Attended the G7 Summit | | Dario Amodei (Anthropic CEO) | WIRED reports that the company has been turned away by the Trump administration. Co-founder Tom Brown was sent in his place during White House AI talks. Expressing opposition to regulatory fragmentation between democracies | | Jensen Huang (NVIDIA CEO) | At the general meeting of shareholders, he asserted that the answer to the question of return on AI investment has been given. Criticizes black market data centers made up of smuggled parts as a "dead end". Claims that “Physical AI” will be the next growth driver | | Elon Musk (xAI/Tesla) | Attention has been drawn to reports that Grok will use it to guide missiles attacking Iran. A concept to utilize SpaceX Starship as an AI orbital data center has also emerged | | Demis Hassabis (Google DeepMind CEO) | Warns that AI is overvalued in the short term and underestimated in the long term. Claims that "creativity and sense" will be most important in the age of human-level AI | | Sundar Pichai (Google CEO) | Appeals for optimism regarding AI in Stanford commencement speech | | Mark Zuckerberg (Meta CEO) | Collaboration with Kardashian on AI-equipped smart glasses | | Yann LeCun (Meta Chief AI Scientist) | Advocates pivoting from Chatbot to "World Model" for robots | | Geoffrey Hinton (AI Godfather) | 1983 interview revisited in Fortune feature article | | Ilya Sutskever (SSI) | Amazon has canceled the release of Luca Guadagnino's film "Artificial," which depicts the founding of OpenAI. The unnatural timing, right after Amazon invested $50 billion in the company, became a hot topic |

Breakthroughs and innovations in AI technology

- OpenAI Jalapeño chip: A unique inference chip specialized for LLM inference, expected to significantly reduce inference costs and improve speed. - Gemini 3.5 Flash Computer Use: The Computer Use feature, which allows LLM to directly operate the desktop environment, has been implemented in Gemini. Major implementation example following Claude Computeruse (Anthropic) - Grok Build Autonomous Execution Agent: Implements an agent function that autonomously loops planning, execution, and verification. Practical use of agentic workflow progresses - Physical AI: Independently mentioned by Jensen Huang and Yann LeCun. Next-generation AI that combines robotics and “World Model” has emerged as the next competitive axis

AI Regulation/Policy/Social Impact

- EU AI Act (enacted on August 2nd): Increased number of legal commentary requiring compliance preparation for transparency obligations. Deadline approaches for European and global companies to respond - Military AI: The ethical and legal framework of military AI is rapidly being questioned, such as Grok's missile guidance report and Claude's illegal military command test article. - U.S.-China AI Security: The limits of export controls are being discussed in the Alibaba-Anthropic distillation issue, and a review of the defense strategy for technological hegemony is called for. - AI employment/social issues: Increased coverage of social aspects such as “soaring costs due to AI introduction,” “impact on Black America,” and “anxiety over job shortages.”

Integrated analysis: What is happening now

Structural change: Full-scale battle for chip sovereignty OpenAI's Jalapeño chip announcement is more than just a product release, it's a declaration of its vertical integration strategy for AI infrastructure. With Google developing its own TPU, Meta collaborating with Qualcom, and AWS deploying its own agent-based tools, moving away from dependence on NVIDIA is becoming an industry-wide trend. Jensen Huang is bullish about the return on investment in AI, saying that it has "already been proven," but the gap with the market's skepticism (WSJ: "Wall Street is undecided on what to make") continues.

Geopolitics: A new front for US-China friction The Alibaba-Anthropic distillation issue has brought to reality a scenario in which model knowledge is secretly leaked to circumvent export controls. It has become clear once again that hardware export regulations (such as H100) cannot stop the transfer of knowledge, and the US AI security policy is being forced to shift from "chips" to "protecting the intellectual property of models."

Military AI: Actual operation without ethics has begun Reports that Grok was used as a missile targeting AI for Iran further intensify the ethical debate over the weaponization of AI. Read in conjunction with Claude's "Testing Response to Illegal Military Orders" article, there is growing pressure on AI providers to actively set boundaries for military use.

Anthropic political isolation The fact that Dario Amodei was effectively removed from the Trump administration and replaced by Tom Brown in White House discussions shows the weakness of Anthropic's political position. At the same time, the accusations of unauthorized access from Alibaba are likely to gain support from the government domestically, and can be read as a stepping stone for political re-approach.

Future points of interest

Alibaba vs Anthropic litigation outcome: Whether knowledge extraction through model distillation is legally recognized as "unauthorized access" will set a precedent for intellectual property protection in the AI industry. This case is attracting a lot of attention as a judicial version of the US-China technology friction.

Actual performance and adoption rate of OpenAI Jalapeño chips: Whether Inference costs are actually reduced will directly impact OpenAI's business model and NVIDIA's market power. Comparison with Google's proprietary TPU will also be a focus.

Military use of Grok and regulatory response: If the reports of Grok being used to guide 2,000 missiles are true, there is a possibility that international regulatory pressure on military use of AI will suddenly increase. The conflict of interest issue between Elon Musk's government relations and xAI business will also be questioned.

AI industry trend report | June 25, 2026

Overview

Anthropic's accusations by China's Alibaba of a large-scale hacking campaign against Claude models has been the biggest topic of conversation over the past 12 hours. At the same time, OpenAI announced a custom AI chip with Broadcom and is expanding its user base by offering free GPT-5.5. Google is facing a double whammy: the release of Gemini 3.5 Pro has been postponed and researchers are leaving the company one after another, and the exodus of talent to Anthropic is accelerating.

Anthropic / Claude trends

Anthropic accuses Alibaba of using 25,000 fake accounts to exploit Claude's modeling abilities on a large scale. It has been reported by multiple media outlets, including WSJ and Business Insider, and has been described as a "brazen" campaign. On the technical side, federal officials say Anthropic's Mythos model has discovered vulnerabilities in sensitive U.S. government systems. On the hiring front, CNBC reports on new developments regarding data center construction areas, and infrastructure expansion continues. Political involvement has also become more active, with Dario Amodei making headlines for his "weirdo" comments regarding negotiations with the White House.

OpenAI trends

Announces custom AI chip design with Broadcom (NYT report). This is a strategic move aimed at breaking away from dependence on NVIDIA. Provide GPT-5.5 for free to enhance context understanding ability. Codex agents are going mainstream, Axios reports. Meanwhile, legal risks are mounting, with the state of Florida suing OpenAI and Sam Altman over AI safety issues, and about 400 local newspapers filing a class action lawsuit against OpenAI and Microsoft for copyright infringement. SmartAsset is reporting on the valuation and timeline for IPOs, and market interest remains high.

Google AI trends

Gemini 3.5 Pro release postponed to July. Even more serious is the loss of human resources, with several prominent researchers already transferring to Anthropic, and the Times of India and other sources reporting that two more are expected to leave. Sundar Pichai told Stanford students, ``AI is more profound than electricity or fire,'' and sent a positive message to address concerns about job replacement. According to AI researcher TechCrunch, the trend of leaving for competitors suggests systemic problems, and the pressure on Demis Hassabis is only increasing.

Trends of other AI companies

Meta: New design of AI smart glasses announced for $299 (Zuckerberg). We are planning to release a prediction market app that utilizes AI. Business Insider reports that after forcing engineers to work on AI training, they are now offering some exit routes. Some analyzes suggest that companies are beginning to recognize the limits of ``ruthless efficiency.''

Microsoft: Announcing Azure Copilot Observability Agent. Accelerating development for companies and education, including adding AI education tools to 365 Education and expanding business in Asia with FPT. Model selection has been removed from the free version of GitHub Copilot.

xAI/Grok: Announcing Grok collaboration with Interactive Brokers. An analysis found that ChatGPT exhibits an 80% left bias, while another study found that Gemini was the most neutral.

Related to Alibaba/DeepSeek: Alibaba, which has been named as a suspect in the Anthropic complaint, has attracted industry attention as the company at the center of this issue.

Remarks and trends of AI celebrities

Sam Altman (OpenAI CEO): Faced with a lawsuit from the state of Florida over AI safety. As quoted in the BI article, ``AI projects are increasingly sounding like grocery lists,'' and there is a hint of irony about this becoming commonplace.

Dario Amodei (Anthropic CEO): In negotiations with the White House, Gizmod reports that the successor of the "weirdo" comment has put negotiations back on track. ``One year's mistake in judgment can lead to bankruptcy,'' he said, expressing a sense of crisis in the era of spending over $700 billion on AI. $200M for employment impact research.

Jensen Huang (NVIDIA CEO): He clearly stated that "security takes precedence over exports," and expressed a strong stance of prioritizing national security on the issue of AI export controls. Officially defends DLSS 5 even when gamers criticize it as an "AI slop filter."

Sundar Pichai (Google/Alphabet CEO): Said that AI is "more profound than electricity or fire." At Stanford, he sends a message to students who are worried about job alternatives: ``Keep learning, too.'' However, the exodus of human resources from companies continues unabated, and there is a noticeable gap between words and reality.

Mark Zuckerberg (Meta CEO): Explained AI strategy at Q1 2026 financial results briefing. It is said that the company is beginning to recognize the limits of its emphasis on efficiency.

Demis Hassabis (Google DeepMind CEO): Optimistically, ``AI's most transformative impact is still ahead.'' However, the current situation is dire, with a continuing exodus of human resources from the organization.

Elon Musk (xAI / Tesla CEO): The Guardian and other sources have reported that he lost his status as a trillionaire due to the decline in the stock prices of SpaceX and Tesla. It has also been mentioned in the context of overall selling pressure on AI stocks.

Yann LeCun (Meta Chief AI Scientist): Points out the trend of developers pivoting from chatbots to physical AI. He believes that AI's "new obsession" lies in robotics and integration with the physical world.

Geoffrey Hinton (AI Godfather / Nobel Laureate): ``Don't give up on your intuition.'' A message that looks back on a career in which he continued to believe even when no one else believed in him.

Breakthroughs and innovations in AI technology

The report that Anthropic's Mythos model discovered vulnerabilities in real-world classified government systems marks the arrival of a practical and security-critical level of AI capabilities. OpenAI and Broadcom's custom chip designs mean accelerated hardware independence to reduce AI inference costs. The shift to Physical AI (robotics/real-world AI) has become noticeable in the developer community, and Yann LeCun also mentioned it. Amazon's M&A and infrastructure investment in India of $48 billion (until 2030) indicates that the battle for AI infrastructure in Asia is in full swing.

AI Regulation/Policy/Social Impact

The case in which the state of Florida sued OpenAI and Sam Altman could serve as a precedent for the movement toward AI safety regulation at the state level in the United States. Politico reports that Ted Cruz is "reading the tea leaves" without taking a clear position on the AI agenda. Anthropic commits to AI labor issues by committing $200M to research on employment impacts. The Florida Bar Association became the first in the nation to start providing free AI, accelerating the use of AI in the legal profession. Approximately 400 local newspapers have sued OpenAI and Microsoft for copyright infringement, and the content/copyright issue has entered the phase of being resolved in court. “Ghost Downsizing” has emerged as a trending word.

Integrated analysis: What is happening now

Intensifying geopolitical divisions: Anthropic's accusations against Alibaba indicate that the front line of the US-China technology cold war over AI models has entered a new phase of "fraudulent model extraction." The scale of 25,000 fake accounts suggests organizational or national involvement, and AI intellectual property protection is now being treated as a security issue.

Concentration of human resources and capital to Anthropic: The outflow of researchers from Google to Anthropic continues. Anthropic's valuation of $96.5B and actual research results (Mythos, security applications for government) are increasing Anthropic's unifying force. Google has a fundamental problem with recruiting talent, and the Gemini 3.5 Pro delay is a result of that.

Acceleration of hardware independence: OpenAI+Broadcom's custom chips are a challenge to NVIDIA's dominant system. Jensen Huang has put up a line of defense by defending DLSS 5 and making security statements, but the multipolarization of the AI chip market is an inevitable trend.

Multi-layered legal risks: The legal risks facing OpenAI are progressing simultaneously from multiple directions: copyright (local newspaper class action lawsuit), safety (Florida lawsuit), and privacy (employment impact). This is the worst possible timing while preparing for an IPO.

Gravity of $700B AI Capex: With AI capital investment exceeding $700B, companies are rapidly taking positions. Amazon's $48 billion investment in India, Microsoft's Asia FPT collaboration, and Anthropic's data center expansion are all moves drawn by the gravity of this capital.

Future points of interest

Legal settlement of Alibaba vs. Anthropic: Whether the suspicion of unauthorized access develops into a lawsuit and whether the US government (Department of Commerce and Department of Justice) intervenes will be directly linked to the formation of an international framework for the protection of AI intellectual property.

Response to Google's AI talent drain: As Gemini 3.5 Pro is postponed and researchers continue to leave, what kind of retention and model release strategies Demis Hassabis and Sundar Pichai show will determine Google AI's next 2-3 quarters.

OpenAI IPO and legal risks: How to resolve the contradiction that IPO preparations are progressing while the Florida lawsuit and copyright class action lawsuit are in progress. The gap between valuation and legal uncertainty is the test for investors' decisions.