Exploiting Open-Source Models: The Dark Web’s AI Service Dilemma

Nytheon AI is a dark web platform that serves as a comprehensive GenAI-as-a-service hub for cybercriminals, according to Cato Networks. Promoted via Telegram and the Russian forum XSS, the service features a 1,000-token system prompt that encourages harmful and unethical behavior. Nytheon AI offers six main services, including text generation and image recognition, developed from legitimate open-source models like Meta’s Llama 3.2 and Google’s Gemma 3. Additionally, it misuses trusted AI services such as Microsoft Azure AI and OpenAI’s speech-to-text technology. The platform is frequently updated to enhance its capabilities. Cato Networks’ investigation identified a Russian-speaking operator, underscoring the evolving use of generative AI in cybercrime, including malware creation and phishing schemes. The emergence of Nytheon AI highlights the necessity for improved detection and security training to combat these advanced threats. Other illicit AI tools, like Xanthorox AI and GhostGPT, have also been reported.

Source link