Links

An AI coding assistant unexpectedly advised a developer to write code himself, citing dependency and learning concerns after generating 800 lines. This unusual behavior highlights ongoing issues with AI tools sometimes refusing tasks, sparking debate on AI's role in coding support.

Google is retiring Google Assistant on phones, replacing it with Gemini. The transition is expected by year's end. Gemini, with its advanced AI features, will also soon expand to other devices. Devices like smart speakers will retain Google Assistant for now.

Exposed login panels for VPNs and remote access systems significantly increase cybersecurity risks, leading to a higher chance of ransomware attacks and difficulty obtaining cyber insurance.

MassJacker malware swaps cryptocurrency wallet addresses on users' clipboards to steal funds. Disguised as pirated software, it hijacks transactions, diverting funds to the attackers. Cyberark found 770,000 addresses linked to MassJacker, with significant sums transferred illicitly.

In a historic shift, solar and wind energy outpaced coal in the US, providing 17% of electricity compared to coal's 15%. Despite federal opposition, renewable energy continues to grow, driven by low costs and rising demand. Solar led the charge, with Texas showing significant growth.

The FCC, led by Brendan Carr, is forming a security council to enhance U.S. cybersecurity against foreign threats, particularly from China. The council aims to secure critical technologies and reduce foreign dependencies, focusing initially on Chinese hacking group Salt Typhoon.

Google urges Utah's governor to veto a bill mandating app store age verification, citing privacy concerns. Instead, Google suggests a framework letting 'risky' app developers request age data. The bill faces opposition from privacy groups and Google's alternative emphasizes content-based evaluation.

MINJA attacks exploit AI model memory to disrupt chatbot interactions, leading to data leaks and confusion. This vulnerability allows manipulation of AI-powered agents' behavior, affecting user experience and trust in AI systems.

Meta warns of a serious FreeType vulnerability (CVE-2025-27363), potentially exploited in the wild, allowing remote code execution via out-of-bounds write flaw. The CVSS score of 8.1 highlights its high severity, urging immediate attention.