AI News Report — 2026-06-28 (Sun)

AI industry trend report | June 28, 2026

Overview

The biggest news is that the Trump administration has partially lifted export restrictions on Anthropic's cutting-edge model, Mythos 5 (Claude Fable 5). At the same time, OpenAI is also restricting the rollout of its next-generation models at the request of the White House, making it clear that the U.S. government will directly intervene in the product rollout of companies at the forefront of AI. DeepSeek announced DSpark, which speeds up Inference by 60-85%, further expanding the cost and speed advantage of Chinese AI companies.

Anthropic / Claude trends

The Trump administration has partially lifted export restrictions on Anthropic's cutting-edge model, Mythos 5 (externally known as Claude Fable 5), and it is expected that supplies to certain U.S. organizations will resume (Fortune, NPR, Business Insider). Gizmodo reports that the ban will be reinstated within a few days, and efforts are underway to completely lift the ban. Meanwhile, AI startups in Asia are taking advantage of the export restriction gap and deploying Mythos-like models, raising the risk that competitors will be the beneficiaries. Anthropic has also published its Economic Index: Cadences report, continuing its analysis of the economic impact of AI on the labor market.

OpenAI trends

Mashable/Barron's reported that the White House is asking OpenAI to limit the rollout of next-generation models. Friction with regulators is hampering product rollout. On the corporate side, there was also talent acquisition news that an executive who led Apple Vision Pro will be transferring to OpenAI (TechCrunch). Valuation risks are also rising, with MarketWatch warning that OpenAI and SpaceX's IPOs could cause a 40% stock market crash. Regarding Sam Altman, there was also a report that Amazon has discontinued distribution of the documentary film (AOL.com).

Google AI trends

Computer Use functionality has been incorporated into Gemini 3.5 Flash, and its practical use as agent-type AI is progressing (Forbes, Yahoo Finance). The deployment of Gemini to Android Automotive (in-vehicle OS) has also been announced, expanding the area in which AI assistants can be embedded. Google.org announces new funding for education and AI partners, and Jack Henry and Google's cloud AI security collaboration strengthens. Sundar Picha gave a keynote speech at India's AI Impact Summit 2026, demonstrating his stance towards global expansion. Regarding the focus on smart glasses, MarketWatch points out the issue of ``How can major tech companies persuade consumers?''

Trends of other AI companies

Microsoft: Satya Nadella promotes trusted 33-year-old executive Jacob Andreou to lead Copilot AI strategy (Fortune, Crypto Briefing). Added new skills and monitoring features to Copilot for Excel, and announced Microsoft 365 Copilot integration with Wix.

xAI/Tesla: Elon Musk said that he is developing "Digital Optimus" with Tesla and xAI. The idea is to use Grok for real-time Computer Action. There are also reports pointing out excessive risks in SpaceX's plan to expand its AI business with a $20 billion loan.

DeepSeek: Announces new Speculative Decoding framework "DSpark". It claims to increase Inference speed by 60-85% per user for DeepSeek-V4 (MarkTechPost). There are also reports that cheap Chinese AI models are expanding their share in the US market, and the Center for National Security (CNAS) has warned of the risk that Chinese AI could counter US military operations.

Meta: It was reported that the UFC (mixed martial arts) has entrusted ranking management to Zuckerberg's AI, but it is "already confused" (lindyssports.com). There was also criticism regarding Meta AI's business model, with concerns that it would lead young people to online gambling.

Remarks and trends of AI celebrities

Sam Altman (OpenAI CEO): A video of Sam Altman sitting with Dario Amodei at the AI Impact Summit 2026 in India but refusing to shake his hand went viral (Mshale). It was also reported that Amazon has canceled a documentary film about him.

Dario Amodei (Anthropic CEO): With the partial lifting of export restrictions on Mythos 5, it is certain that Anthropic will resume deployment of its flagship model. Business Insider, Euronews, etc. reported all at once.

Elon Musk (xAI/Tesla CEO): Comments on the sharp rise in AI chip prices as "the biggest price increase we've ever seen" (CryptoRank). Announced the Digital Optimus project, indicating the direction of integrating Grok as Tesla's behavioral AI. WSJ reported on the existence of big-name investors who are not betting on Musk's AI vision, but against it.

Jensen Huang (NVIDIA CEO): ``The world is coming to me. I just have to wait.'' This statement was featured on Benzinga and attracted attention as a symbol of NVIDIA's dominant market position. His name has also been mentioned in South Korea's Saemangeum investment bid.

Sundar Pichai (Google/Alphabet CEO): Delivered a keynote speech at India AI Summit 2026, highlighting Google's involvement in global expansion in Asia.

Mark Zuckerberg (Meta CEO): Issues have emerged in the social evaluation of Meta AI due to reports of failure in adopting UFC ranking AI and criticism of promotion of gambling among young people.

Demis Hassabis (Google DeepMind CEO): Talked about the future of AI and creativity on "Think with Google" and outlined DeepMind's research direction.

Yann LeCun (Meta Chief AI Scientist): Mentioned in an article pointing out concerns that large models in the United States are moving towards closed systems in the name of "security" (ChainCatcher).

Breakthroughs and innovations in AI technology

DeepSeek's DSpark is the biggest Inference technology news. It is a framework that uses speculative decoding to improve the throughput of DeepSeek-V4 by 60-85%, once again demonstrating the power of open source Chinese AI research. Qualcomm announces plans to port data center-grade AI processing power to smartphones (24/7 Wall St.). In the medical field, the combination of AI and minimal residual disease (MRD) detection attracted attention at the ASCO Breakthrough 2026 conference. Google is accelerating edge AI deployment with Gemini integration into Android Automative.

AI Regulation/Policy/Social Impact

The biggest policy news is the partial lifting of Anthropic's Mythos 5 export restrictions. The Trump administration is torn between regulation and industrial protection, and the government continues to intervene directly in the OpenAI model rollout (Politico). The Hill warned that "AI is creating a new underclass in America." The Guardian argued that "the AI bubble will expand even further before it bursts," and MarketWatch presented a 40% crash scenario due to the OpenAI/SpaceX IPO. In the legal field, Scale Law Firm AI has appointed a leader specializing in AI education and training, and AI penetration into the legal world is progressing.

Integrated analysis: What is happening now

The biggest theme running through the current AI industry is ``the normalization of US government intervention''. The two events, Anthropic's model export restrictions and OpenAI's request to limit rollouts, signal a new era in which AI companies' product roadmaps are driven by political decisions. This fundamentally changes the risk profile of AI companies.

In terms of power, the rise of DeepSeek continues, and DSpark's acceleration of Inference has further solidified the "cheap and fast Chinese AI" narrative. The more access to American-made models is restricted due to export regulations, the more customers from Asia and the Global South are flowing to DeepSeek and other products.

As for technological trends, there is a clear shift to the practical application phase of agent-type AI (Computer Use). Use cases in which AI autonomously operates computers are entering the industrial implementation stage, such as Gemini 3.5 Flash's Computer Use integration, the Digital Optimus project using Grok, and Interactive Brokers' ChatGPT/Grok integration.

On the investment front, the mainstream media is beginning to spread warnings about a bubble over overheated investment in AI infrastructure, and the market sentiment on the eve of the IPO rush is an even balance between risk-on and risk-off.

Future points of interest

Full recovery timing for Claude Fable 5 (Mythos 5): Full recovery is expected within days of partial release, which will test the scale of market share that Anthropic can regain against the competition.

OpenAI's next-generation model and the White House's compromise: The extent to which government intervention continues and how OpenAI negotiates will determine future product roadmaps.

DeepSeek DSpark empirical data: If the claim of 60-85% inference speedup is supported by independent verification, the balance of cloud inference cost competition may be significantly disrupted.

AI industry trend report (June 28, 2026)

Overview

WSJ reports that China has reached the same level as Anthropic in the field of AI cybersecurity, and the composition of the race for AI supremacy is changing. OpenAI jointly developed and introduced its own chip "Jalapeño" with Broadcom, accelerating vertical hardware integration. On the other hand, a situation occurred where Google restricted Meta's use of Gemini, and the strain on the infrastructure for AI demand became apparent.

Anthropic / Claude trends

WSJ has reported that Chinese AI has caught up to Anthropic's cybersecurity standards, and some are calling it "AI Blip 2.0." Anthropic's Fable 5 model is said to be close to receiving clearance from the US government (Axios). There are signs of a change in political positioning both internally and externally, with reports that Dario Amodei has been removed from MTG in the White House and replaced by his co-founder. In conjunction with OpwnAI's expected postponement of its IPO to 2027, developments surrounding Anthropic's Mytos 5 (Mythos 5) are also attracting attention.

OpenAI trends

It is reported that the Jalapeño AI chip with Broadcom was developed and introduced in nine months, accelerating the hardware independence strategy. Uber hires Prabhjeet Singh for Indian market to strengthen Asian expansion. Financial information was leaked before the IPO, and it was reported that losses last year amounted to $39 billion. The state of Florida files a lawsuit against Sam Altman and OpenAI over AI safety. GPT-5.6 (Sol, Terra, Luna) was announced, but the number of users who could use it was initially extremely limited.

Google AI trends

FT reports that Google has set a cap on the usage of Meta's Gemini model. The main reason for this is said to be a tightening of computing capacity due to the rapid increase in demand for AI. The trade-off between external provision to competitors and in-house demand has become apparent, and the issue of cloud AI supply constraints may spread throughout the industry. In news related to Google DeepMind's Demis Hassabis, reports have also been confirmed that OpenAI has restricted the release of new models after undergoing a US security review.

Trends of other AI companies

Meta: Approximately 8,000 layoffs were reported due to Zuckerberg's large-scale AI investment shift. Coupled with the unusual situation of Google restricting access to Gemini, we are forced to reconsider our AI strategy.

Microsoft: Added Finance Skills, Planner, and external data link functions to Excel Copilot. Copilot has been quietly reinstalled on Windows 11 and continues to infiltrate endpoints.

xAI/SpaceX: The US announced that Grok was used in a military strike against Iran, marking the first official confirmation of military AI use. There are also reports that SpaceX will acquire the coding tool Cursor for $60 billion, aiming to enter the enterprise coding market.

DeepSeek: DSPark released in collaboration with Peking University. This release symbolizes the shift from parameter competition to cost competition in large-scale models, improving inference efficiency by 60-85%.

Remarks and trends of AI celebrities

Sam Altman (OpenAI CEO): A video of Sam Altman meeting Prime Minister Modi with Dario Amodei (Anthropic CEO) at the Delhi AI Summit 2026 has gone viral. It is also mentioned that ``I only become conscious of Wall Street after reaching $500 million in sales.''

Jensen Huang (NVIDIA CEO): ``AI will become the control plane of quantum computers'' and NVIDIA announced the Ising platform. Admitted that revenue from China had dropped to zero and presented an alternative plan worth $20 billion. Claims that AI will help increase manufacturing jobs in Texas.

Dario Amodei (Anthropic CEO): There are reports of alternative responses to the White House meeting, and there may be a change in the way political influence is exercised.

Elon Musk (xAI/Tesla CEO): It was reported that he made remarks that could be seen as suggesting a SpaceX-Tesla merger. He is reported to have issued the same warning to the world along with Tim Cook, and interest in AI safety continues to be intertwined with business strategy. Tesla hits major AI chip milestone and expands beyond cars.

Demis Hassabis (Google DeepMind CEO): 36Ki reported that RSI (Recursive Self-Improvement) will arrive in 2028.

Ilya Sutskever (SSI Co-Founder): Name mentioned when Amazon movie about Sam Altman also passed on Netflix and Focus.

Geoffrey Hinton (AI Godfather): A critique was published in 36Kan stating that the heated advertising battle surrounding Anthropic's Distillation is nothing more than a "fight for Narrative Dominance."

Breakthroughs and innovations in AI technology

DeepSeek's DSPark increases inference speed by up to 85%, setting a new standard in efficiency for LLM inference. Liquid AI has released LFM2.5-230M, increasing the practicality of on-device inference with support for llama.cpp, MLX, vLLM, SGLang, and ONNX. The OpenAI-Broadcom joint Jalapeño chip has attracted industry attention for its unprecedented development speed of nine months, and the custom silicon strategy of foundation model companies is accelerating. DeepSeek's movements indicate a structural shift in which ``the end of parameter competition and the rise of cost competition'' can be seen.

AI Regulation/Policy/Social Impact

In the United States, Republican Rep. Moran has proposed a bill that would require AI safety reporting. There are also reports that the Trump administration will restrict access to advanced AI models, complicating the export control framework. With the official confirmation of Grok's use of Military Strike, the policy debate surrounding the weaponization of AI is expected to further accelerate. The Bank for International Settlements (BIS) warns of the risk that "excessive excitement" about AI could lead to the bursting of a long-term investment bubble. 97,000 job cuts were announced in May 2026 in the US, 40% of which will be caused by AI.

Integrated analysis: What is happening now

Changes in the balance of power: The rise of Chinese AI (DeepSeek's DSPark, competing in the cybersecurity field) suggests that the US and China are moving from a bipolar conflict to a phase of convergence in technological levels. While Anthropic's Fable 5 clearance is said to be near, Dario Amodei's access to the White House is changing, with political positioning emerging as a competitive factor.

Infrastructure strain exposed: The fact that Google has restricted Gemini offerings to Meta shows that AI demand is already exceeding top companies' infrastructure supply. This means that the transition to an "AI-first economy" may be faster than expected, and at the same time there will be serious bottlenecks on the supply side.

Acceleration of hardware vertical integration: OpenAI-Broadcom's Jalapeño, Tesla's AI chip milestone, NVIDIA's Ising platform and quantum collaboration declaration. A "layer dissolution" is underway, with software companies descending to the silicon layer and hardware companies ascending to the AI control plane.

Military/Geopolitical AI Applications: Grok's use of Military Strike has historical significance as the first officially confirmed example of AI being explicitly integrated into the decision-making loop of a military system. The impact on future regulations and international law will be immeasurable.

Future points of interest

Anthropic's ability to obtain Fable 5 clearance: The timing and conditions of US government approval will determine the speed of Anthropic's expansion into the Enterprise and Gov markets.

OpenAI's Jalapeño chip mass production and IPO strategy: The decision to postpone the IPO amid the development of proprietary silicon and a $39 billion deficit will determine the overall temperature of the AI investment market in 2026-2027.

Trends in international regulation of military use of AI: Following the official confirmation of the use of Grok, the focus will be on whether moves to formulate guidelines for the use of AI weapons at the UN, NATO, and individual government levels will take shape.