Something that the AI industry hasn't understood yet (because most people in AI ignore the law): From a legal perspective, general-purpose AI systems (like ChatGPT) don't make sense, as mandatory guardrails and product liability claims cannot be properly allocated. A product that can tell bed stories for kids and ALSO create erotica-style engagement with adults will never be adequately regulated, as the people affected and the risks are different, and the rules for each use case should be different. In non-AI products, erotica and toys are never coupled in the same device. It makes no sense and in many cases it would be illegal. In AI products (like AI systems) they should probably be separated as well. Specialized AI systems might be powered by general-purpose AI models (like LLMs), but the fine-tuning, mandatory guardrails, and liability rules of each system should be compatible with the type of risks and people affected. I don't see a future where AI is properly regulated and general-purpose AI systems exist. Hopefully most countries will realize that soon.