Anthropic has confirmed a security incident in which attackers gained access to active Claude user sessions, prompting the company to forcibly sign out affected accounts and delete stored payment card information. The response, first reported by Notebookcheck, came after evidence emerged that session tokens had been stolen and used to access user accounts without authorization.
What Happened and How Anthropic Responded
Session hijacking attacks work by stealing authentication tokens that keep users logged in, bypassing the need for passwords entirely. In this case, affected Claude users found themselves suddenly signed out of their accounts. Anthropic also cleared saved credit and debit card details as a precaution against unauthorized purchases or subscription changes. The company has not publicly disclosed how many accounts were affected or how the tokens were initially stolen.
Key Facts
- Claude user session tokens were stolen by unknown attackers
- Anthropic force-logged out all affected user accounts
- Saved payment card information was wiped from compromised accounts
- The method used to steal session tokens has not been publicly disclosed
- No official count of affected users has been released
The incident follows a pattern of security issues that have surfaced around the Claude platform in recent months. Earlier this year, concerns arose when Claude users' private conversations appeared in Google Search results, suggesting data handling practices were not as contained as users expected. That episode sparked significant user backlash and questions about how conversation data is stored and exposed.
Session token theft is one of the most effective attack vectors against web applications because it sidesteps authentication entirely. Once an attacker has a valid token, they are the user as far as the server is concerned.Security researcher commentary via Notebookcheck
A Broader Security Picture
This incident is unlikely to exist in isolation. Security researchers have previously flagged the Claude ecosystem as an active target for malicious actors. A separate campaign involved fake Anthropic websites being used to distribute infostealer malware targeting Claude Code users. Infostealers are precisely the kind of tool used to harvest session cookies and authentication tokens from infected machines, making that campaign a plausible upstream source for incidents like this one.
For Anthropic, the timing is awkward. The company has been on an aggressive growth trajectory, and a series of security incidents could undermine user trust at a critical moment. Wiping payment cards was the right call from a risk management standpoint, but it also signals that the company believed the threat to financial data was credible enough to act on immediately rather than investigate first.
Users who were affected are advised to change passwords, enable multi-factor authentication if not already active, and review any recent account activity for unauthorized changes. Anthropic has not yet issued a detailed public post-mortem on the incident, which security advocates say is needed to allow users to assess their own exposure.
The broader question the incident raises is how session tokens were obtained in the first place. If the source was malware on user devices, the problem lies outside Anthropic's direct control. If tokens were exposed through a server-side vulnerability or an insecure API, the responsibility sits squarely with the company. Until Anthropic provides more technical detail, users are left guessing about their actual risk level. Those following the latest Claude AI news should watch for any further disclosure from the company in the coming days.
