A ModelScope MS-Agent vulnerability allows attackers to feed malicious commands to AI agents and modify system files or steal ...
Malicious AI browser extensions collected LLM chat histories and browsing data from platforms such as ChatGPT and DeepSeek.