According to @timnitGebru and as reported by 404 Media, the widespread use of AI agents powered by large language models (LLMs) is undermining traditional security protocols and frameworks developed over decades. The article highlights a case where users granted extensive permissions to LLMs, allowing unrestricted access and control, which exposed critical vulnerabilities, such as in the Moltbook database incident. This trend raises significant concerns about security best practices in enterprise AI adoption, emphasizing the urgent need for new frameworks that address the unique risks of LLM-based agents.