Researchers Unveil AI Worms Capable of Spreading Across Systems, Raising Cybersecurity Concerns

As generative AI systems such as OpenAI’s ChatGPT and Google’s Gemini continue to advance, they are increasingly being integrated into various applications, from automating mundane tasks like calendar bookings to more complex operations. However, this growing autonomy also introduces new security risks. A group of researchers has recently demonstrated the creation of what they claim are the first generative AI worms capable of spreading from one system to another, potentially stealing data or deploying malware.

Why it Matters

The development of these AI worms represents a significant shift in cybersecurity threats. Traditional malware and viruses typically require human intervention or specific vulnerabilities to propagate. In contrast, AI worms can autonomously identify and exploit vulnerabilities across connected systems, expanding the attack surface exponentially. Ben Nassi, a researcher at Cornell Tech and one of the creators of the worm, emphasizes this point: “It basically means that now you have the ability to conduct or perform a new kind of cyberattack that hasn’t been seen before.”

Key Risks

The AI worm, dubbed Morris II in homage to the original Morris computer worm from 1988, was tested in controlled environments. The researchers demonstrated how it could attack a generative AI email assistant to steal data from emails and send spam messages, bypassing some security measures in both ChatGPT and Gemini. This capability underscores several key risks:

1. Data Theft: AI worms can exfiltrate sensitive information, including personal and corporate data, from compromised systems.
2. Malware Deployment: These worms can deploy additional malware, further compromising the target system and potentially creating backdoors for future attacks.
3. Spam and Phishing: By leveraging generative AI to create convincing phishing messages, these worms can trick users into divulging sensitive information or installing malicious software.

The Opportunity

While the creation of AI worms presents significant risks, it also highlights opportunities for enhancing cybersecurity measures:

1. Advanced Threat Detection: The development of AI-driven threat detection systems can help identify and mitigate AI worm attacks more effectively.
2. Enhanced Security Protocols: Tech companies and startups must implement robust security protocols to prevent AI systems from being weaponized. This includes regular vulnerability assessments and patch management.
3. User Education: Educating users about the risks of AI worms and best practices for protecting their data can reduce the likelihood of successful attacks.

Current State and Future Outlook

As of now, generative AI worms have not been observed in the wild. However, multiple researchers agree that this is a security risk that should be taken seriously by the tech industry. The increasing multimodal capabilities of large language models (LLMs), which can generate text, images, and even video, further complicate the threat landscape.

Nassi and his team’s research, shared exclusively with WIRED, provides a critical early warning for the cybersecurity community. By understanding the potential risks and working proactively to develop countermeasures, stakeholders can better protect against this emerging threat.