Follow
"We Are Still Unable to Secure LLMs from Malicious Inputs"
https://www.schneier.com/blog/archives/2025/08/we-are-still-unable-to-secure-llms-from-malicious-inputs.html
