First signs of AI-enabled ransomware attacks emerge

The day when artificial intelligence (AI) will be used to autonomously launch ransomware attacks is now one step closer. ESET security researchers report they have discovered an AI ransomware tool on VirusTotal, a service for analyzing files and URLs to detect malware, that can decide on its own whether to exfiltrate, encrypt or even destroy data.

Dubbed PromptLock, the tool uses a freely available language model from OpenAI to generate malicious Lua scripts on the fly to execute various tasks. Since it was discovered, a group of professors, research scientists and PhD students at NYU Tandon School of Engineering have come forward to confirm they uploaded the prototype, which they dubbed Ransomware 3.0, to VirusTotal during testing.

The NYU team published a research paper describing how the malware prototype they developed as a proof of concept uses large language models (LLMs) to autonomously plan, adapt and executive a ransomware attack. While PromptLock has yet to be discovered being employed in the wild by cybercriminals, it may be a harbinger of a new wave of attacks launched by machines rather than human adversaries.

The NYU research paper comes on the heels of a research paper from Carnegie-Mellon University (CMU) that previously demonstrated how LLMs can also be taught to autonomously plan and execute real-world cyberattacks against enterprise-grade network environments.

Defending against AI-enabled attacks

The implications of all this research for cybersecurity professionals is profound. It’s now more a question of when, rather than if, cybercriminals will set up a ransomware service based on agentic AI technologies that leverage LLMs to identify vulnerabilities, determine how best to exploit them, create the required code and orchestrate the actual attack in a matter of minutes. Cybersecurity teams that are already finding it difficult fending off cyberattacks will soon be overwhelmed by waves of cyberattacks that will cost adversaries very little to launch.

Of course, the only way to fight AI fire is going to be with AI. Organizations, ready or not, are going to need to invest in AI platforms that can respond at machine speed to the attacks being launched by other machines. Without that capability, a successful attack will soon be able to wreck untold amounts of havoc in a few seconds. Cybersecurity professionals are simply not going to be able to respond fast enough to limit the scope of the blast radius of a breach without relying on AI to detect and thwart attacks in real time.

The challenge, as always, is finding the funding needed to acquire the AI tools and platforms that will be required. There is little doubt that cybersecurity budgets will continue to increase, but given how fundamentally the threat landscape is about to change, they may not growing fast enough to keep up. That may not be an easy conversation to have with business and IT leaders, but facts are facts. After all, as cybersecurity professionals well know, the only thing worse than spending more money on cybersecurity is a cyberattack that paralyzes the business to the point where no revenue is being generated at all.