AIThe Registerabout 3 hours ago
Researcher poisons open-weight AI model for under $100
3 min read
Katie Paxton-Fear, a cybersecurity lecturer at Manchester Metropolitan University and Semgrep staff security advocate, installed a backdoor in an open-weight AI model in about an hour for less than $100. Only ten training examples made the model's code output reliably vulnerable to remote code execution. Larger models were easier to poison.
Level
Hype check
Tap to vote and see what everyone thinks.
Summary by ByteBrief
