A one-prompt attack that breaks LLM safety alignment

Large language models LLMs and diffusion models now power a wide range of applications, from document assistance to text-to-image generation, and users increasingly expect these systems to be safety-aligned by default. Yet safety alignment is only as robust as its weakest failure mode. Despite...

Processes may panic upon receipt of malicious protobuf messages

Issue Details A security issue was discovered in code generated by the gogo protobuf compiler used by Kubernetes. The gogo protobuf compiler issue has been assigned CVE-2021-3121 and is also known as the “skippy peanut butter bug”. A program which uses affected code to handle a malicious protobuf...