Apple's New Siri Becomes an "Everything Tool" — Available in the iOS 27 Public Beta

Apple's dramatically revamped Siri is evolving from a mere voice assistant into the core of the iPhone user experience. According to WIRED, the new Siri is already available to try through the iOS 27 public beta.

While the previous Siri was limited to recognizing a narrow set of commands, the new Siri is designed as an "Everything Tool" that bridges every kind of action on the iPhone. Thanks to enhanced Apple Intelligence, cross-app coordination, contextual understanding, and multi-step task execution have become considerably more natural.

This direction embodies an industry-wide trend of smartphone AI assistants being integrated at the OS level. Google's Gemini and Samsung's Galaxy AI take similar approaches, but Apple's strength lies in tight hardware-software integration.

MIT Unveils New Method to Detect Improper Training in AI Models

MIT has announced a new method for detecting whether an AI model was trained on child abuse images (CASM) — without generating any images.

The technique identifies the presence of improper training data by analyzing the model's internal representations. The crucial point is that detection works without actually generating harmful content, and it is regarded as a practical breakthrough in AI safety research.

Auditing training data for large-scale models has become an industry-wide challenge, and this technology is expected to contribute to regulatory compliance and model safety assurance.

Meta's Next-Gen AI Glasses Can Start the Camera Without an Indicator Light

It has been reported that Meta's next-generation AI glasses, currently under development, are designed so that the camera can be activated without lighting the indicator lamp. According to Privacy Guides, privacy concerns are mounting.

Because wearable AI devices by their nature keep a camera and microphone on your body at all times, privacy-focused design is critically important. Recording without an indicator means people nearby have no way to notice they are being recorded, making this an issue that calls for social consensus.

AI Infrastructure Investment Adds Inflationary Pressure — AP Points Out

According to AP reporting, the rapid construction boom of AI data centers has become a new inflation threat. It has been flagged as an issue that could even influence the Federal Reserve's policy decisions.

From electricity demand and construction materials to the scramble for semiconductor talent, the AI buildout is having ripple effects across the entire economy. While some argue that AI-driven productivity gains will ease inflation in the long run, the analysis is that the construction boom acts as upward price pressure in the short term.

Fujitsu Develops Self-Evolving Multi-Agent Platform "MAAF"

Fujitsu has developed "MAAF," a multi-AI-agent platform for business operations, and has begun verifying it. The system automatically configures itself from data such as meeting recordings and safely self-evolves based on its operational history.

Rather than a "build it and be done" approach to AI agents, this continuous-improvement model points to the next step in enterprise AI operations. Fujitsu aims to drive company-wide AI adoption by integrating the platform with its own AI foundation.

Government AI "源内" Built on Domestic Models Set to Go Live

A "purely domestic government AI" leveraging Japan-developed AI models built by NTT and others, along with domestic cloud infrastructure, is set to start operating for the Digital Agency's government AI, 源内. Digital Minister Matsumoto said it will be "a pioneering initiative."

Building a government AI that reduces dependence on foreign models and ensures data sovereignty and security marks an important step in Japan's digital strategy.

---