AI Daily News — 2026-06-29 (Mon)

AI industry trend report | June 29, 2026

Overview

The biggest topic of conversation is the geopolitical movement in which Anthropic's Claude Fable 5 has been revived from a two-week export ban by the US government, and Europe is moving to attract Anthropic. xAI announced Grok 4.5 as a V9 model with 1.5 trillion parameters and started internal testing at Tesla and SpaceX. On the other hand, the lack of AI supply is emerging as a cross-industry issue, with Google restricting Meta's use of Gemini.

Anthropic / Claude trends

The biggest focus is on lifting export restrictions on Claude Fable 5. There have been a number of reports that the US government has approved the revival of Fable 5 after a security review. During the ban period, Austria was lobbying the EU to host Anthropic in Europe, making it clear that US AI export restrictions are having the opposite effect of promoting European independence. On the technical side, headlines read, ``Claude Opus 4.8 solves AI's biggest problems,'' and there are reports that Claude AI has achieved 20 times the performance of the robotics team. The Bulletin of the Atomic Scientists discusses why high-performance models like Claude Fable and Mythos don't fit within traditional export control frameworks, and the legal vacuum in AI regulation has sparked debate.

OpenAI trends

Focus on hardware strategy and capital policy. OpenAI and Broadcom's joint announcement of the first custom AI chip to compete with NVIDIA sent ripples through the semiconductor industry. There are also reports of negotiations with Qualcomm regarding smartphone AI chips, and OpenAI's desire for vertical hardware integration is accelerating. Regarding the IPO, an article appeared stating that SpaceX is considering postponing its IPO due to stock price fluctuations. In terms of human resources, it was reported that Apple Vision Pro's VP position was transferred to OpenAI, and OpenAI continues to have an advantage in the competition to acquire talented personnel. Sam Altman continues his global expansion by speaking at the AI Impact Summit in India.

Google AI trends

Tight supply and demand symbolized by supply constraints and Gemini access restrictions. The Financial Times reported that Google had restricted Meta's Gemini API access, and multiple media outlets followed suit and confirmed this. The reason is a lack of capacity, which clearly shows that the demand for AI exceeds the supply, and even Google can't keep up. It was revealed that six researchers from DeepMind's Coding department were transferred to Meta, OpenAI, and Anthropic, and the impact on Alphabet stock is being discussed. There have also been articles criticizing Demis Hassabis' management decisions, and DeepMind's talent loss problem is serious.

Trends of other AI companies

xAI: Announced Grok 4.5 as a V9 model with 1.5 trillion parameters. Elon Musk himself announced the start of private betas with Tesla and SpaceX. It has also been revealed that the FTC has approved Musk's acquisition of AI networking startup "Mesh". Meanwhile, Yann LeCun's public statement that xAI is a failure and Musk is having a hard time hiring top talent has caused a stir within the industry.

DeepSeek: Announced that it released "DSpark" that utilizes Speculative Decoding and increased the inference speed of V4 models by up to 85%. Chinese media has positioned this as the "next DeepSeek moment," suggesting a response to the U.S. technology blockade.

Microsoft: Enhances Copilot's capabilities for Excel, adding reusable skills and data connectors. On the other hand, Edge's AI history search function has been abolished. There are also warning reports that prompt injection is emerging as an attack targeting enterprise AI RAG pipeline agents.

Meta: Appeared as a person who received gemini access restrictions from Google. Employee tracking program suspended after internal data breach. There have been reports of ``snacks being provided'' as a measure against declining morale, and articles that suggest management turmoil are prominent. There is also a critical report (WION) stating that ``Meta's AI has destroyed the lives of 2 million people.''

Remarks and trends of AI celebrities

Sam Altman (OpenAI CEO): He spoke at the AI Impact Summit in India and demonstrated his stance on global AI deployment. GIGAZINE features a feud between Altman and Dario Amodei that has spanned over 10 years.

Elon Musk (xAI / Tesla CEO): Announced the start of private testing of Grok 4.5 in the form of a personal leak. There are also reports that SpaceX has invested $60 billion in AI, and the question ``Are we building the next Amazon?'' has become a hot topic in the market. SoftBank CEO Masayoshi Son questioned Musk's vision, asking, ``What is the significance of an AI data center in space?''

Jensen Huang (NVIDIA CEO): ``No company can match NVIDIA's cost performance.'' We remain bullish even though NVIDIA stock is down 18% in 2026. He expressed the view that the market environment where demand for AI exceeds infrastructure supply will work in NVIDIA's favor.

Dario Amodei (Anthropic CEO): It attracted the dual attention of Fable 5's unbanning and EU invitation. Technical results regarding Claude AI's robotics performance were also reported.

Yann LeCun (Meta Chief AI Scientist): Declares xAI a "failure" and points out Musk's recruitment difficulties. He also mentioned the shift of top developers to physical AI and appealed for a shift from AI development centered on chatbots.

Sundar Pichai (Google/Alphabet CEO): Faced with two challenges: Gemini's supply and demand problem and DeepMind's talent outflow. An article in which a former Google employee talks about the "crazy traits" that Larry Page, Sergey Brin, and Sundar Pichai have in common attracted attention.

Mark Zuckerberg (Meta CEO): Social criticism of Meta's AI services is increasing, with critical articles calling for ``Zuckerberg to resign'' appearing.

Demis Hassabis (Google DeepMind CEO): The decision to pivot to the coding department has been criticized for causing an exodus of researchers, and an essay has even appeared calling for demotion or firing.

Breakthroughs and innovations in AI technology

Grok 4.5 (xAI): The 1.5 trillion parameter V9 model is larger than the GPT-4o model and indicates the direction of larger scale MoE architecture. DeepSeek DSpark: As a practical implementation of Speculative Decoding, it is noteworthy that it has achieved a significant reduction in inference cost (up to 85% speedup). In the edge AI field, Qualcomm is making a full-fledged move into the AI data center field to compete with NVIDIA, and along with OpenAI and Broadcom's joint development of custom chips, the multipolarization of the semiconductor ecosystem is accelerating. Reports that Apple iOS 27 will allow users to select third-party AI models signal an important shift in the openness of mobile AI.

AI Regulation/Policy/Social Impact

The most important topic is Claude Fable 5 export regulations and European movements. The precedent set by the U.S. government by temporarily banning the export of high-performance AI models for security reasons and then reinstating them will have a major impact on the framework of future AI export regulations. The fact that the EU, led by Austria, is considering bringing Anthropic to Europe is a sign of geopolitical competition over "digital sovereignty" in AI. Pope Leo XIV's declaration that ``AI cannot be morally neutral: Ethics is embedded in all algorithms'' is another notable religious and philosophical position statement. The issue of AI copyrights in entertainment companies remains unresolved and is expanding, and there is also an example of the FBI using AI to investigate the attack on the White Palace Banquet Party.

Integrated analysis: What is happening now

Supply constraints, geopolitics, and hardware independence are converging as three major themes. The situation in which Google is forced to restrict Meta's use of Gemini shows the reality that cloud AI supply is not keeping up with demand. This is completely consistent with Jensen Huang's assertion that demand is outstripping supply.

Anthropic's Fable 5 ban/unban episode exposed weaknesses in US AI governance. Europe's decision to take advantage of this opportunity to host a bid could accelerate competition between the United States and Europe regarding AI governance.

xAI's Grok 4.5 announcement and DeepSeek's DSpark show that players other than OpenAI/Anthropic are seriously catching up in both high-performance models and efficient inference. In particular, DeepSeek's inference cost reduction approach serves as an alternative to costly proprietary models.

Movement to break away from dependence on NVIDIA is also clear. OpenAI+Broadcom's custom chips, Qualcomm's entry into data centers, and Apple's rising device prices (due to a shortage of AI chips) show that the reorganization of the semiconductor ecosystem is progressing as a real demand.

Human resource mobility is also an important signal. The movement of human resources between companies, such as the exodus of researchers from DeepMind to Meta, OpenAI, and Anthropic, and Apple executives changing jobs to OpenAI, is causing a reallocation of technological capabilities.

Future points of interest

International expansion of Claude Fable 5 and establishment of a permanent framework for export controls — Whether the move to attract Europe comes to fruition or the US institutionalizes export controls may be decided in the coming weeks.

Assessing the ability of OpenAI + Broadcom custom chips and Qualcomm's strategy to compete with NVIDIA — Whether or not it can compete with NVIDIA's cost performance at the actual product level could fundamentally change the industry's dependence on semiconductors.

The impact of Google DeepMind's talent outflow and Gemini supply constraints — How Alphabet resolves the double whammy of internal organizational collapse and external supply shortages will determine the competitiveness of Google AI.

# AI industry trend report (last 12 hours) June 29, 2026

Overview

Anthropic's ``Claude Mythos 5'' has passed the US government's review and is headed for a limited release, while xAI's Grok 4.5 has entered internal testing and Musk has suggested that it will outperform Claude Opus, and the race for frontier models is rapidly accelerating. In an unprecedented situation, Google restricted Gemini API supply to Meta due to a lack of compute, and the AI infrastructure bottleneck also caused a rift in the relationship between Big Tech. In China, the number of API calls has ranked first in the world for nine consecutive weeks, and the US share has plummeted from 72% to 33%, and the geopolitical balance of AI power is undergoing a major change.

Anthropic / Claude trends

Telecompaper reported that the US government has approved the limited release of ``Claude Mythos 5''. Access is only available to selected organizations, and it appears that military and security purposes are also envisioned. According to The Information, the presence of Mythos shook DeepSeek and helped the company raise $7.4 billion in funding. Spring Tools 5.2.0 adds experimental support for the Claude Code plugin, which is making its way into the developer ecosystem. In a 10,000-word interview, Dario Amodei talks about how to balance business and safety as AI becomes a superweapon, and Austria is moving to attract Anthropic to Europe.

OpenAI trends

GPT-5.6 "Sol" was announced as a cutting-edge model specialized in cybersecurity (SecurityWeek). At the same time, it has been reported that the general release of GPT-5.6 has been delayed due to the US government's request for early access. Sam Altman told Korean media, ``We will create fully autonomous AI researchers by March 2028,'' giving a concrete timeline for reaching AGI. The ecosystem continues to expand, including a strategic partnership with HP Inc. and the transfer of Apple Vision Pro executives to the OpenAI hardware team. Meanwhile, SoftBank's stock price hit a one-month low due to uncertainty regarding the OpenAI IPO.

Google AI trends

The most notable is the report that ``Google has restricted the supply of Gemini API to Meta'' (Gizmodo, The Indian Express, Fortune India). It is said that the compute capacity could not keep up with Meta's demand and the supply had to be capped, and a competition for compute between Big Tech has come to light. Meanwhile, Gemini is being installed in Renault's vehicle fleet, and it was reported that the company is also developing full control functions for Android smartphones. It has also announced Connected AI Tools for the education field, expanding its penetration into both B2C and B2B.

Trends of other AI companies

xAI/Grok: Musk reveals SpaceX/Tesla internal testing of Grok 4.5 and suggests that it is "already comparable in performance to Claude Opus." The company also announced plans to release one large model a month this year.

Meta: New AI research chief declares "agents are the next real-world milestone" despite Google's limited supply of Gemini. It was also reported that the company plans to move much of its content moderation to AI.

DeepSeek/China: China ranks first in the world in the number of AI model API calls for 9 consecutive weeks. While Nvidia's chip sales to China are stagnant, domestic chip manufacturers such as Huawei are emerging (Washington Post).

Microsoft: Stock price rises 6% in one day due to AI stock rebound. We are expanding our deployment in regulated industries by publishing case studies of Copilot for law firms such as Johnson Stokes & Master.

South Korea: The Ministry of Science and Technology announced a quadrillion-won investment in an AI data center.

Remarks and trends of AI celebrities

Sam Altman (OpenAI CEO): Declaration with a specific deadline: ``We will realize fully autonomous AI researchers by March 2028.'' The country is also facing poll results showing that voters of both parties want stricter regulation of AI.

Elon Musk (xAI/Tesla CEO): Popular topic as he celebrates his 55th birthday. Internal testing of Grok 4.5 has begun and comments suggesting it will "exceed Claude Opus" have spread to the market. There is also analysis that mesh optical communication infrastructure (Starlink/xAI integration) will be the main battleground in the next AI infrastructure war.

Dario Amodei (Anthropic CEO): In a major interview, he talked about the tension between safety management and business growth as AI becomes a "super weapon." Expansion into Europe (attracting Austria) is also taking shape.

Sundar Pichai (Google CEO): The difficult management decision to limit Gemini supply to Meta has come to light. This inadvertently exposed the structural problem of a lack of compute.

Jensen Huang (NVIDIA CEO): Mentioned regarding the announcement of investment in AI data centers in South Korea. The headwind of sluggish chip sales in China continues.

Yann LeCun (Meta Chief AI Scientist): He has caused controversy within the industry by saying via MSN that ``Musk's xAI is a failure'' and ``There is an AI bubble.''

Demis Hassabis (Google DeepMind CEO): Speaking on "The Future of AI and Creativity" on Think with Google. Application of the World Model concept to AI has emerged as a major theme.

Geoffrey Hinton (AI Godfather/Nobel Laureate): Warns against data center construction boom: "Without an AI plan, it will be too late."

Breakthroughs and innovations in AI technology

Attention to AI agents is rapidly increasing. Meta's new research chief has declared that "agents are the next real-world milestone," and tech media has even run a special feature titled "Let's talk about AI agents." The World Model is attracting attention as a concept that is spreading from psychology to the main field of AI, and Demis Hassabis is increasingly being mentioned in this context. AI security and reliability tools are also on the rise, such as the Deepfake detection model "Halo" announced in partnership with Qualcomm at Computex 2026. The cybersecurity-specific GPT-5.6 Sol is also a product in this context. AI chip design using a thermodynamic approach (Thermodynamic AI Chip) was featured on Machine Learning Street Talk, and innovations in the hardware layer continue.

AI Regulation/Policy/Social Impact

In the US, a poll published by NBC News shows that voters of both parties support stricter regulation of AI, showing signs of a political consensus forming. A framework for the US government to control access to the most powerful models (Techzine Global) is also in motion, and the limited release of Claude Mythos 5 is a concrete example of this. The issue of compatibility between AI and net-zero (carbon neutrality) has also emerged on the policy agenda. UN University pointed out that ``AI alone cannot achieve the SDGs; policies need to catch up.'' In the legal industry, the ``end of the billable hour'' model is being discussed on Bloomberg Law, and the transformation of the industry structure through AI is becoming a reality. Countries such as South Korea, Austria, and Kazakhstan have announced their own strategic investments in AI one after another, and AI sovereignty is becoming established as a new keyword.

Integrated analysis: What is happening now

Multipolarization of frontier model competition: OpenAI (GPT-5.6 Sol), Anthropic (Mythos 5), and xAI (Grok 4.5) are competing for release in parallel, and the former "OpenAI domination" has shifted to a completely multipolar structure. Musk's declaration of "one large model per month" may further accelerate this competition.

Compute From Bottleneck to Geopolitical Weapon: Google's cap on Meta's API usage shows that compute shortages have gone from being just an engineering problem to being a bargaining card between companies and nations. The stagnation of Nvidia's chips for China and the rise of domestically produced chips in China have combined to make the geopolitical division of AI infrastructure clear.

China's rapid catch-up: Ranking first in the world in API calls for 9 consecutive weeks, and the US market share plummeting from 72% to 33% is a symbolic figure. The report that DeepSeek was shocked by the existence of Mythos 5 and raised emergency funds is proof that the Chinese are seriously pursuing the frontier model.

Declaration of transition to the age of agents: Multiple celebrities and companies have agreed that “agents are next”, and the practical application of agentic AI will be the main battleground in the second half of 2026.

Future points of interest

Performance comparison between public release of Grok 4.5 and Claude Opus: Whether or not Musk's statement of ``exceeding Claude Opus'' is proven by actual benchmarks will greatly influence the market evaluation of xAI.

Expansion of the scope of limited release of Claude Mythos 5 and US access control policy: ``Rights of access to models'' that determine which countries and organizations can use it may become a new geopolitical diplomacy card.

What will become of Google-Meta's compute battle: The balance of power between open source and closed model will change depending on whether Meta reduces its dependence on Gemini and concentrates on LLaMA, or whether Google increases compute and resumes supply.