SN 1099: An Alien Mind - From RSA Weakness to Agentic Attacks
Summary
Security Now episode 1099 explores the release of GLM 5.3, a powerful open-weight AI model that rivals proprietary frontier systems in autonomous vulnerability discovery and exploit generation. The hosts discuss the implications of "restraint obliteration," which allows users to remove safety filters from open models, and critique Anthropic’s efforts to restrict access to advanced cyber capabilities. Additional topics include a surprising reduction in RSA crypto strength, new Firefox vulnerabilities, and the growing challenges of AI alignment as models become harder to monitor.
Listen ad-free on Castria