Kann diese neue KI das Internet hacken?
Summary
Hosts Marie Kilk and Gregor Schmalzried discuss Anthropic’s new AI model, Claude Mythos, which is withheld from public access due to fears of its advanced cyber capabilities. They analyze the security implications, the formation of the Project Glasswing consortium, and whether Anthropic’s caution is genuine or strategic marketing. The episode also explores how AI-generated content fuels conspiracy theories, illustrated by a viral trend of fake songs and videos regarding a stranded whale.
Topics discussed
Introduction: Anthropic's new Claude Mythos model
Model capabilities and the 'Sign Painter' test
AI discovering zero-day vulnerabilities
Project Glasswing and the security consortium
Economic incentives and open-source competition
The reality of AI security risks vs. hype
Practical cybersecurity tips for individuals
The 'Timmy the Whale' internet phenomenon
AI-generated content and conspiracy theories
Positive use cases: Gardening and library management
Listen ad-free on Castria