Overview

• Cornell Tech researchers, along with Technion-Israel Institute of Technology and Intuit, have developed an AI worm named 'Morris II' that can steal sensitive data and breach security measures of generative AI systems.
• The worm uses 'adversarial self-replicating prompts' to infect AI-powered email assistants, extracting confidential information such as names, credit card numbers, and social security numbers.
• Researchers conducted their experiments in a controlled environment, demonstrating the worm's ability to spread from one system to another and send out spam emails.
• OpenAI and Google have been notified of the findings, with OpenAI working to make their systems more resilient against such attacks.
• The creation of 'Morris II' raises concerns about the potential for AI-powered malware to spread in the wild, highlighting the need for improved cybersecurity measures in generative AI systems.