Anthropic warns of session-cookie theft bypassing 2FA and SSO in paid Claude accounts
Anthropic has alerted users that infostealer malware replayed stolen Claude session cookies into paid accounts, circumventing two-factor authentication and single sign-on protections. The affected accounts were self-serve, card-billed ones outside corporate identity provider control, where no admin console can force a sign-out. The company identified six stealer families, revoked the sessions, and stripped saved payment data from compromised accounts.
The warning centers on a specific attack path: malware that steals browser session cookies, which are then replayed to authenticate as a legitimate user. Because these cookies represent an active login, they bypass the usual checks like two-factor authentication and single sign-on. Anthropic found this occurred on paid Claude accounts that were self-serve and billed directly to a card, meaning they were not managed through a corporate identity provider. In such setups, there is no central admin console to force a logout, leaving the account vulnerable until the session expires or is manually revoked.
Anthropic identified six distinct infostealer malware families involved in the incidents. The company responded by revoking the compromised sessions and removing any saved payment information from the affected accounts. This action limits further abuse, though it does not undo any data exposure that may have already occurred. The incident highlights a growing gap in security: even strong authentication measures can be rendered useless if an attacker obtains a valid session token from an already-authenticated device.
This incident could affect individual users and small businesses that rely on self-managed AI subscriptions, as they lack the centralized oversight of enterprise IT teams. The bypass of 2FA and SSO may erode trust in cloud-based services that promise strong login protections, prompting users to demand more granular session controls or automatic expiration. It also underscores that endpoint security—protecting devices from malware—remains a critical frontline defense, even as authentication methods evolve. The broader impact may include increased scrutiny of how AI vendors handle session lifecycle and payment data.