A soft-fluff factory is dangerous if there aren't safeguards. It's not that ai or llms are dangerous as a concept. It's that they are being given authorities and access that they shouldn't as there are inadequate safeguards built in and there are inadequate punishments for the neglectful development. Make the companies and users liable for the errors and they will work hard in that rather than chasing investment and PR with all their effort.

Its often the case with new technology that the law doesn't match how they are used. However, as technology progresses, thinks are scaled and rolled out at a much more rapid pace. Hence the term hyperscalers. The law needs to catch up. Instead the law seems to give concessions. Partly due to regulatory capture, partly due to economic reasons, partly due to lack of understanding, partly due to corruption.