Cybercriminals are Seeking AI Experts to Develop Malicious Language Models
Cybercriminals are on the lookout for AI and machine learning experts to create custom, malicious large language models (LLMs) after existing tools failed to meet their advanced intrusion needs, according to security researchers.
Underground forums are buzzing with discussions about exploiting the guardrails set by AI-powered chatbots from companies like OpenAI and Google-owned Gemini. Etay Maor, senior director of security strategy at Cato Networks, noted that hackers are exploring ways to develop malicious LLM products.
One Russian-speaking threat actor named Poena was observed recruiting experts on Telegram for this purpose. Ransomware and malware operators are also following this trend, given the surge in demand for AI talent.
Existing tools like WormGPT, XXXGPT, and XXX WolfGPT have fallen short of meeting the demands of cybercriminals, leading them to seek AI experts for developing private GPTs to break existing restrictions and create malicious models.
Recorded Future highlighted in a report that threat actors are using generative AI to develop malware and exploits, aiming to evade detection tools used by LLM applications. This includes modifying malware source code to evade detection using AI-powered LLM systems.
As cybersecurity measures improve, such as limiting activities like generating malicious code and crafting phishing emails, private GPTs are becoming more vulnerable to data leaks. Security experts suggest implementing more preventive measures and using multilayered malware detection capabilities to combat the use of AI by threat actors.
While access to AI resources remains a constraint for lower-tier cybercriminals, advanced nation-state actors are likely to use AI for more sophisticated attacks, such as creating deepfakes and spreading disinformation. Overall, the use of AI in cybercrime continues to evolve, raising concerns among security experts about the need for enhanced preventive measures.
