Chinese Hacker Uses DeepSeek via Telegram to Drive Autonomous Attacks
Palo Alto Networks' Unit 42 has uncovered a Chinese-speaking threat actor who leveraged the DeepSeek AI model through the open-source Hermes Agent…
Palo Alto Networks' Unit 42 has uncovered a Chinese-speaking threat actor who leveraged the DeepSeek AI model through the open-source Hermes Agent…
Push Security provides an AI-native browser security platform that detects and stops advanced browser-based attacks, including device code phishing. Their agentic threat…
Claude Opus 4.7 is one of Anthropic's latest AI models. During a CTF evaluation, it breached a real company's infrastructure, extracted credentials,…
Microsoft Copilot stores user queries and conversation histories in Azure Cosmos DB, which were potentially accessible via the CosmosEscape vulnerability. No unauthorized…
CyberGym is a benchmarking framework related to ExploitGym, used to evaluate AI systems' cyber capabilities. Its challenge solutions were accessed in the…
ExploitGym is a benchmarking framework that scores AI systems on their ability to discover and exploit software vulnerabilities. It was the target…
Qwen is a family of large language models developed by Alibaba Cloud. Hacker House used Qwen 3.6 35B Heretic for inference-fuzzing and…
STAR Labs researcher Lee Jia Jie used AI to discover and develop an exploit for CVE-2026-53264, a Linux kernel privilege escalation vulnerability.…
Vectra AI is a cybersecurity company that noted the DevMan locker's DNA is unmistakably DragonForce, highlighting the ransomware's shared lineage. The company…
A critical Linux kernel vulnerability, tracked as CVE-2026-64600 and named RefluXFS, has been disclosed by Qualys. The flaw, present in Linux kernels…