Swiss AI Models Found Vulnerable to Security Breaches

EPFL researchers reveal critical security flaws in major AI language models, achieving 100% success rate in bypassing safety measures, raising concerns about AI regulation in Switzerland.