LLM eat a lot of energy and are their hallucination are bad: [Wayback/Archive] LLM-generated passwords ‘fundamentally weak,’ experts say • The Register
Your AI-generated password isn’t random, it just looks that way
…
AI security company Irregular looked at Claude, ChatGPT, and Gemini, and found all three GenAI tools put forward seemingly strong passwords that were, in fact, easily guessable.
…
Basically they are almost as good as the 2007 XKCD “four” number generator, the 2013 XKCD “I’m So Random” or the 2001 Dilbert “nine” number generator further below (don’t read the latter if you dislike Scott Adams)
Is it a coincidence or are these two using two small squared numbers?
Anyway: avoid LLM whenever possible, as most often they do more bad than good.
And for passwords, better use the blog post that was already scheduled for tomorrow: Generating random strings for passwords and uuids/guids on both Windows and Linux using base64 and hex encoding, plus: “Hive Systems: Are Your Passwords in the Green?”
Via [Wayback/Archive] Eloy.: “LLMs are centrist randomness: not useful for anything that requires truth but neither for password generation” – HSNL Social
Below this post, there are some great responses as well.
The Wiert Corner - irregular stream of stuff 