12 articles
As AI technology rapidly advances, so do its malicious applications. This article delves into the escalating global threats posed by AI misuse and analyzes how future efforts in technological innovation, policy-making, and international collaboration can build robust defenses to ensure AI's positive development.
OpenAI confirmed its ChatGPT service refused to assist an individual linked to Chinese law enforcement in planning an online campaign to discredit the Japanese prime minister. This incident not only highlights the robust safety mechanisms of AI models but also underscores AI's escalating role in information warfare and geopolitical conflicts.
PyTorch has disclosed a high-severity vulnerability (CVE-2026-24747) with a CVSS score of 8.8. Attackers can exploit specially crafted models to achieve Remote Code Execution (RCE), even when 'safe loading' options are enabled.
Security researchers are warning that the ecosystem surrounding OpenClaw, the self-hosted AI assistant, has become a hotbed for malware. Last month, 14 malicious 'skills' were discovered on ClawHub, specifically designed to drain cryptocurrency wallets.
Google DeepMind and the UK AI Security Institute (AISI) have announced a deepened partnership focused on critical AI safety research. This collaboration ensures early access to frontier models for independent safety evaluations, targeting risks in cybersecurity, biology, and autonomous behavior to ensure responsible development.
Google DeepMind has released Gemma Scope 2, a comprehensive interpretability suite for the Gemma 3 model family. Utilizing Sparse Autoencoders (SAEs), it provides a 'microscopic' view of neural activations, marking a breakthrough as the first open-source tool to support multimodal AI interpretability across text and vision.
At CES 2026, Nvidia's Vera Rubin chips and Alpamayo platform signal the dawn of Physical AI. Meanwhile, xAI's Grok faces global backlash over nonconsensual deepfakes, prompting landmark legislation like New York's RAISE Act as the world grapples with AI safety.
Nvidia's acquisition of Groq for $20 billion signals a new era of 'inference-optimized' compute. This report covers the Meta-Manus deal, Cursor's expansion, Micron's HBM shortage, and the latest in AI safety regulation for early 2026.
Equipped with NVIDIA DRIVE AV technology, the Mercedes-Benz CLA achieved the highest safety score in the 2025 Euro NCAP assessments, signaling a global shift from passive crash protection to AI-driven accident prevention.
OpenAI introduces gpt-oss-safeguard, a family of open-weight reasoning models specifically optimized for safety classification, allowing developers to implement and iterate on custom security policies locally.
As AI agents gain the ability to browse the web autonomously, security becomes a paramount concern. OpenAI has detailed its defense architecture focused on preventing URL-based data exfiltration and prompt injection, ensuring sensitive data isn't leaked to malicious servers.
OpenAI has announced the €500,000 EMEA Youth & Wellbeing Grant to support NGOs and researchers in studying the impact of AI on youth safety, mental health, and development. Part of the EU Economic Blueprint 2.0, this initiative highlights the tech giant's commitment to social responsibility and the digital wellbeing of the next generation.
