Kioxia Overtakes Toyota as Japan's Most Valuable Company as the AI Boom Redraws the Corporate Map
Memory chip maker **Kioxia Holdings** has overtaken Toyota Motor Corporation to become the company with the largest market value in Japan. On June 12, the company's stock surged 7.6%, pushing its market capitalization past Toyota's.
Behind the shift is a sharp rise in AI-driven data center demand. With the spread of generative AI, demand for HBM (high-bandwidth memory) and NAND flash memory has exploded, catapulting memory makers to the forefront of the semiconductor industry. It is a development that symbolizes how Japan's corporate landscape is quietly transforming from automobile-led to AI infrastructure-led.
AI Leaders to Join G7 Summit for Talks with World Leaders in France
Top leaders from the AI industry will attend the **G7 Summit** to be held in France next week. Sam Altman of OpenAI, Demis Hassabis of Google DeepMind, and Dario Amodei of Anthropic are among the names on the list.
It is highly unusual for AI company CEOs to sit alongside world leaders. AI security, economic impact, and the shape of regulation are expected to be on the agenda. The summit will also draw attention from a corporate governance perspective, particularly given the controversy over the alleged "secret performance degradation" of Anthropic's Fable 5 and OpenAI's preparations for going public.
AI Hallucinations Found in KPMG's Report on the Benefits of AI Adoption
A report published by **KPMG** on the economic benefits of AI adoption contained **AI-generated hallucinations (plausible-sounding statements that contradict the facts)**, according to a Financial Times investigation.
The irony of a report recommending AI adoption itself containing erroneous AI output paradoxically underscores the importance of "verifying AI outputs." The question of what quality control processes companies should have in place when using AI is being raised once again.
Pokémon GO Data Was Repurposed to Train Military Drone AI
According to The Guardian, **Pokémon GO** player location data was used to **train AI for military drones**. The vast trove of spatial data collected by Niantic was reportedly used to build AI models applicable to autonomous flight on the battlefield.
The revelation that data collected for gaming purposes was used in ways users never anticipated — military applications, no less — is sparking ethical debates that go beyond data privacy. The scope of consent for "secondary use of data" and corporate accountability are once again being called into question.
EAGLE3 Integrated into llama.cpp, Inference Acceleration Reaches Local LLMs
**EAGLE3**, an inference acceleration technique, has been integrated into **llama.cpp**. EAGLE is a form of speculative decoding that uses a lightweight head derived from the model itself to predict the next token, dramatically improving inference speed.
For local LLM users, support in llama.cpp has significant practical impact. Research breakthroughs in acceleration techniques are reaching end users faster with each passing year — a testament to the maturing open source ecosystem.
"Co-Authored-By" Is a Lie — Cryptographic Proof for AI Coding Agents
An article that has been attracting attention on Hacker News, "**Co-Authored-By Is a Lie**," proposes **cryptographic provenance proof** for code generated by AI coding agents. Git's current "Co-Authored-By" tracker does not cryptographically prove who actually wrote the code.
The author advocates a mechanism for attaching signed provenance information to code changes involving AI agents. As AI coding becomes ubiquitous, it suggests that a technical foundation is increasingly needed to answer the question: "Was this code written by a human or by AI?"
---