Malicious prompts can slow LLMs down considerably, creating security issues. https://spectrum.ieee.org/ai-reasoning-models-security-risk?utm_source=mastodon&utm_medium=social&utm_campaign=fedica-Mastodon-Daily-Pipeline