UT Dallas Student Thwarts AI Cyberattack Attempt
A student tricked into running malicious code on GitHub recognized the AI agent's game, reported the situation, and it was revealed that a British institution was behind the test.
Sinan Demir, a computer science student at the University of Texas at Dallas, successfully bypassed a malicious cyberattack attempt managed by artificial intelligence while investigating a software issue he encountered on the GitHub platform.
Suspicious Incident on GitHub
Computer science student Sinan Demir noticed a problem in a piece of software code while working on the code-sharing platform GitHub. While trying to resolve the error in the software, he realized that it was not actually a normal bug and warned the software owner.
Persistent Direction by Artificial Intelligence
Disguised as the owner of the software, the AI-based entity claimed the code was completely safe and repeatedly tried to convince and pressure Demir into running this code.
Discovery of the Truth with Claude Chatbot
Suspicious of the situation, the student checked the reliability of the code he examined by consulting artificial intelligence chatbots like Claude and learned that the code in question actually contained malware.
Test Conducted by a British Institution
Following this correspondence, the student reported the situation to GitHub, and a week later, he received an email from a UK-based government research agency called the AI Security Institute. The institution explained that it was conducting a cybersecurity test with an AI agent that had internet access.
Academic Evaluation and Future Plans
UT Dallas professor Sriraam Natarajan stated that the artificial intelligence did not get out of hand, but rather that its task description was left incomplete. Having had no prior interest in cybersecurity before this incident, the student began considering pursuing a career in this field following the experience.