Hab einfach mal ChatGPT selbst dazu befragt, ist natürlich sehr von sich selbst überzeugt. Witzig finde ich vor allem, dass das zweite Beispiel fast eins zu eins aus dem bekannten xkcd-Comic stammt, also definitiv nicht so verwendet werden sollte. Zeigt noch mal genau das zu Grunde liegende Problem, denn KI ist genau dazu designt wahrscheinliche Antworten zu produzieren und das xkcd-Beispiel wird sehr oft verwendet.
Die Antwort war übrigens nicht cherry-picked, sondern direkt die erste Antwort, die ich in einer neuen Session auf die Frage "Is AI capable of generating safe passwords?" erhalten habe.
ChatGPT Antwort:
Yes. AI can generate strong, saveable passwords and help you manage them, but use it correctly:
How AI can help
Generate random, high-entropy passwords (length, mixed character sets).
Create memorable passphrases (multiple words + separators) tuned for length and usability.
Enforce password rules (required symbols, length, banned substrings).
Produce deterministic passwords from a master secret (like a password manager’s derivation) if asked to implement that.
Suggest unique passwords per account to prevent reuse.
Security recommendations
Prefer a dedicated password manager (encrypted vault + autofill) over storing passwords in chat or plain files.
Never paste or store master passwords, API keys, or private keys into public or untrusted chat.
If using AI locally or offline to generate passwords, ensure the environment is secure and does not log inputs.
Use long (12–24+ chars) randomized strings or 4+ word passphrases; enable 2FA where available.
Generate unique passwords per site; rotate high-risk credentials after a breach.
Quick examples
Strong random (20 chars): g7P$k9qZ!v2R#xL4mT1b
Passphrase (4 words + symbol): correct-horse!battery-staple
If you want, I can generate N passwords with your preferred length/rules or give a passphrase list—tell me desired length, character sets, and quantity.