TechCrunch Maps Session Checks for ChatGPT, Claude, Perplexity

A practical walkthrough from TechCrunch's Lorenzo Franceschi-Bicchierai treats consumer AI accounts the way security columnists have long treated inboxes and social profiles: assume a determined attacker eventually knocks on the door, and learn where the 'who's currently logged in' panel lives before you need it. The piece covers ChatGPT, Claude, and Perplexity, and the through-line is that each vendor has tucked the useful controls into a slightly different corner of settings.

For ChatGPT, the path is account settings, username, Security and Login, then Active Sessions, where a user can boot suspect devices one by one or hit 'Log out all.' Password resets require signing out first, then using 'Forgot password' to trigger a six-digit code. Claude puts the same list under Settings then Account, with a three-dot menu to end a session and a 'Log out of all devices' option; because Claude uses email-based login links rather than a password, there is no password to rotate. Perplexity is the outlier here, since it does not expose individual sessions at all, so a worried user's only real move is 'Sign out of all sessions' and a fresh six-digit code sent by email to log back in.

The baseline advice under all three is the familiar pairing of a password manager and multi-factor authentication, and Franceschi-Bicchierai notes that ChatGPT and Perplexity both offer MFA while Claude's email-link flow is a different beast. Reading something this basic in 2026 still earns its space because chat histories, saved instructions, and whatever files or context a user has piped into these tools have quietly turned an AI login into one of the more revealing things on a person's device.

The piece stops short of tying itself to any specific breach or campaign, and it does not say whether the three vendors send a login-alert email when a new device signs in, which is arguably the control that would matter most. Perplexity's missing per-session view is called out but not pressed on, and there is no mention of how Enterprise or Team tiers handle session revocation differently from consumer plans. That gap aside, it is the kind of reference page a security lead can hand to non-technical colleagues who now spend their working day inside these three products. It lands in the middle of a long run of AI security coverage here, 298 stories in the last 90 days alone.