Study Warns AI Chatbots Leaving Users with Dangerously Weak Passwords

(MENAFN) Cybersecurity researchers have sounded the alarm over a growing blind spot in digital security — AI-generated passwords that look robust on the surface but harbor critical vulnerabilities underneath, according to new findings from cybersecurity firm Irregular.

ITPro reported Thursday that Irregular's analysis of leading chatbots — including ChatGPT, Claude, and Google Gemini — found these tools consistently produce predictable, patterned outputs rather than the genuinely random strings that secure passwords demand.

The results were stark. Repeated outputs and severely limited variation plagued test results, with one Claude-generated password appearing 18 times across trials. Irregular flagged "strong regularities" in GPT-5.2's behavior, noting that "nearly all passwords begin with a v, and among those, almost half continue with Q."

The implications become even more alarming when measured against industry standards. Password security is quantified through entropy — a measure of unpredictability — and while a properly secured 16-character password should register approximately 98 bits, AI-generated equivalents averaged a deeply insufficient 27 bits, rendering them significantly more vulnerable to attacks, researchers said.

Kevin Curran, a senior member of the Institute of Electrical and Electronics Engineers (IEEE) and cybersecurity professor at Ulster University, did not mince words, labeling dependence on AI for password creation a "risky practice."

"These models often produce strings, which appear strong and complex but are actually highly predictable," he said, adding they "do not generate true randomness."

In response, experts are urging users to pivot toward dedicated password managers, passkeys, and multi-factor authentication as more reliable lines of defense.

MENAFN22022026000045017169ID1110773269