How to check if your ChatGPT, Claude, or Perplexity account was hacked

Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- TechCrunch published a guide for checking whether accounts on ChatGPT, Claude, and Perplexity have been compromised, recommending unique passwords stored in a password manager plus multi-factor authentication where available.
- ChatGPT offers multi-factor authentication and a password-based login; users can view logged-in devices under Settings → Security and Login → Active Sessions and reset their password via a six-digit code emailed to them.
- Claude does not use passwords at all — Anthropic sends a login link to the user's email — and there is no MFA option, meaning the only recourse against a suspected breach is terminating active sessions from Settings → Account.
- Perplexity does not display where users are currently logged in, offering only a blanket "Sign out of all sessions" button under All settings, with re-login handled through a unique six-digit code sent by email.
- The guide highlights that only ChatGPT and Perplexity support MFA, while Claude's email-link authentication model leaves users without the second-factor protection available on the other two platforms.
- Each platform exposes a different level of session transparency: ChatGPT shows devices individually with a single-device logout option, Claude allows per-session termination via a three-dot menu, and Perplexity offers only an all-or-nothing sign-out.
Why it matters: The guide exposes meaningful security gaps that users should weigh when choosing an AI assistant: Claude offers no MFA and no password to change after a suspected breach, while Perplexity does not even display active login locations, leaving users with a blunt all-or-nothing sign-out as their only option. For the millions using these tools for sensitive work, the practical ability to detect and respond to a compromised account varies sharply by platform — ChatGPT users get the most granular controls.
Ask SkimNews




