Trump asks the impossible of Anthropic

Gary Marcus argues that guardrails for LLMs are fundamentally impossible, citing security experts and his own 2024 warnings. He concludes that the issue is not specific to Anthropic but a general Generative AI problem.