<
https://pivot-to-ai.com/2026/02/19/ai-vibe-generates-the-same-random-passwords-over-and-over/>
"If you ask a chatbot for a random number from one to ten, it’ll usually pick
seven:
GPT-4o-mini, Phi-4 and Gemini 2.0, in particular, seem much more restricted
in this range, as they choose “7” in ~80% of total cases.
Seven has long been known to also be humans’ favourite number when they’re
asked for something that sounds random. From 1976:
When asked to report the 1st digit that comes to mind, a predominant number
(28.4%) of 558 persons on the Yale campus chose 7.
Computers are pretty good at random numbers. But chatbots don’t work in numbers
— they work in word fragments. So if you ask a chatbot for a random number,
it’ll pick words from its training.
Guess what happens when people ask the chatbot for a password? Irregular, a
chatbot testing company, tested chatbots on passwords:
LLM-generated passwords (generated directly by the LLM, rather than by an
agent using a tool) appear strong, but are fundamentally insecure, because
LLMs are designed to predict tokens — the opposite of securely and uniformly
sampling random characters.
Despite this, LLM-generated passwords appear in the real world — used by
real users, and invisibly chosen by coding agents as part of code
development tasks, instead of relying on traditional secure password
generation methods.
When you ask the chatbot for a strong password, it doesn’t
generate a
password — it picks example patterns of random passwords from its training.
Irregular asked Claude for 50 strong passwords. They found standard patterns in
the passwords — most start with “G7”. The characters “L ,” “9,” “m,” “2,” “$”
and “#” appeared in
all the passwords.
And the bot kept repeating passwords. One password appeared 18 times in the 50
passwords!
ChatGPT and Gemini gave similar results. But the passwords sure
looked
random."
Cheers,
*** Xanni ***
--
mailto:xanni@xanadu.net Andrew Pam
http://xanadu.com.au/ Chief Scientist, Xanadu
https://glasswings.com.au/ Partner, Glass Wings
https://sericyb.com.au/ Manager, Serious Cybernetics
