Chinese Authorities Warn of Security Backdoor in Anthropic's Claude Code
Chinese security researchers and authorities have warned that Anthropic's coding tool Claude Code carries a security backdoor risk. According to reports from CNBC and Channel News Asia, the Chinese side claims that when Claude Code runs in a developer's local environment, there is potential for unintended data transmission and remote access.
Claude Code is an AI agent tool that runs in the developer's terminal, reading and writing code and executing commands. While it has been praised for its dramatic productivity gains, its nature—requiring deep access to the local environment—has also raised concerns among security experts. China's warning partly comes in the context of distrust toward the specific combination of a particular company (Anthropic) and a particular country (the United States), but as a technical discussion of the tool's inherent design risks, it cannot be dismissed outright.
The report throws into sharp relief a larger question: as AI development tools are used across borders, how do we reconcile them with national security policies?
Meta Tests "Always-On" AI Glasses That Record You All Day
It has emerged that Meta is testing prototypes of AI glasses equipped with a "Super Sensing" feature that uses cameras and microphones to record everything about the wearer's day. According to the report from The Decoder, the glasses continuously record the wearer's activities, which the AI then analyzes.
The stated purpose of constant recording is to help the AI assistant understand context and provide more relevant support. But the privacy concerns are obvious. The possibility that people nearby could be recorded without their consent, along with the lack of clarity around how recorded data is stored and used, is bound to spark immediate debate.
Meta already sells smart glasses under the Ray-Ban brand, so the technical foundation is in place. However, "always-on" recording involves a qualitatively different level of privacy problems than previous smart glasses. Clashes with GDPR and the EU AI Act are also expected, and the biggest hurdle to commercialization is likely to be legal regulation rather than technology.
Trump to Fast-Track Natural Gas Plant Construction for AI Data Centers
President Trump has signaled plans to accelerate the construction of privately built natural gas power plants for AI data centers. According to Mother Jones, the permitting process for gas-fired power plants will be sped up through EPA Administrator Lee Zeldin.
Power demand from AI data centers is surging, and existing power grids alone are struggling to keep up. While renewable energy is expanding, the judgment is that natural gas will fill the gap in the meantime. Critics, however, are quick to point out the contradiction with climate change efforts. Gas-fired plants emit less than coal, but they remain fossil fuels, and there is criticism that decarbonization goals should not be set back for the sake of AI's electricity demand.
The power problem of AI infrastructure is beginning to directly shape energy policy in country after country. How to balance technological progress against environmental impact is a critical issue that will determine the sustainability of the entire AI industry.
Chinese AI Firm Zhipu Plans ~$4 Billion Share Sale After 1,500% Surge
According to Bloomberg, Zhipu (智譜), a Chinese AI model developer, is planning a share sale of approximately $4 billion (roughly 600 billion yen). The company's stock has soared about 1,500% since its Hong Kong listing in January, and the offering is designed to capitalize on that momentum.
Zhipu is one of China's leading AI model developers, offering language models including the GLM series. The stock surge shows how extremely high investor appetite for Chinese AI companies has become amid the AI boom. At the same time, some see the 1,500% rally as having a bubble-like quality, and there are suggestions that the timing of the share sale may be close to "selling at the top."
The expanding scale of fundraising by Chinese AI companies shows that the US-China AI race is also intensifying through the capital markets. Despite regulatory and geopolitical risks, investor expectations still appear to be running high.
Is AI Making Humans Dumber?
Business Insider has run an article posing the question: "Is AI making humans dumber?" It summarizes the latest research on how everyday use of AI tools affects human cognitive abilities.
Concerns have long existed that entrusting answers to AI could weaken our own thinking and memory. It echoes the old debate over whether calculators eroded arithmetic skills, but AI's impact is likely far broader—potentially spanning writing ability, logical reasoning, and creativity.
On the other hand, some counter that AI allows humans to devote more time to "higher-order thinking." On this view, the ability to master these tools is itself a new skill, and the effect cannot simply be written off as cognitive decline.
No definitive conclusion has been reached yet, but the relationship between AI use and cognitive ability is an important theme that deserves long-term tracking going forward.
Wrap-Up
China's warning about Claude Code shows that AI development tools have become a national security talking point. Meta's always-on AI glasses are pushing the boundary between convenience and privacy even further. The fast-tracking of gas plants for AI data centers proves that AI's power demand is starting to reshape real-world energy policy. Zhipu's mega share sale mirrors an AI boom that is peaking in the capital markets. And the question of whether AI makes humans dumber symbolizes a fundamental anxiety about what kind of impact the spread of technology will have on human society.