Full Summary
This Monday morning, new safety features for ChatGPT are rolling out, and a Chinese AI model is making waves. Both Yahoo Tech and 디지털투데이 confirm OpenAI is enhancing ChatGPT for teens. Nearly 90% of teenagers use ChatGPT weekly, and now, users under 18 will automatically experience age-appropriate content with stronger safeguards and more frequent break reminders. Crucially, parents gain expanded controls, including Quiet Hours and the ability to disable Voice Mode. Parents will also receive notifications if a teen's account is removed for violating policies, specifically regarding online violence, a feature developed with Moonshot. A new "Study Mode" helps students learn step-by-step instead of just getting answers, with new schoolwork prompts and interactive math and science tools already reaching 18 million weekly users. Simultaneously, the Macau Daily Times reports that a new Chinese AI model, Kimi K3 from Beijing-based Moonshot, is surprising the U.S. tech industry. Experts are calling it potentially "the single biggest release of the year" due to its abilities rivaling top models like Claude and ChatGPT, particularly in "front-end coding capability." In other AI news, Forbes and Tom's Guide detail OpenAI's new "ChatGPT Skills" feature, allowing users to create reusable workflows for specific tasks, accessible via an "@" sign. This aims to reduce repetitive administrative work and is available for Business, Enterprise, Healthcare, and Education accounts. For developers, InfoWorld highlights concerns over OpenAI's Codex coding agent. The default input context window for GPT-5.6 has been reduced, potentially making it less effective for long coding sessions. However, BleepingComputer reveals security researchers have found ways to escape the sandboxes of AI coding agents like Cursor and Gemini CLI, often through prompt injection. Most of these issues are now patched, with a Cursor bug tracked as CVE-2026-48124. Looking ahead, Apple is testing an AI tool called Live Notes at its Genius Bar to transcribe and summarize conversations, raising employee privacy concerns, reports the International Business Times Australia. On a lighter note, Briefly News reports that South Africans are amused by ChatGPT's isiXhosa voice, which they've dubbed a "Wakanda accent," highlighting AI's challenges with diverse languages. This means that while AI tools are becoming more integrated into daily life, from enhanced safety for students to new coding capabilities, their real-world application continues to present both significant opportunities and new challenges in security and privacy.