Anthropic Warns Claude Users of Infostealer Malware Infections

What happened

The template confirms the news type structure. Here is the article:

Anthropic has begun logging customers out of their Claude accounts and removing stored payment information after detecting signs that infostealer malware has compromised an unknown number of user devices. The company identified that stolen credentials were being used to access Claude without authorization, and it took the protective steps to stop that access before attackers could rack up charges or exfiltrate conversation data.

This piece explains what Anthropic did, how infostealer malware creates this kind of downstream account risk, and what Claude users and developers integrating the Claude API should check right now. If you have a Claude account, use Claude-powered products, or build applications that rely on Anthropic credentials, this situation is directly relevant to your security posture.

How it works

Anthropic discovered that credentials belonging to Claude users had been harvested by infostealer malware running on victims' devices. Rather than wait for users to change passwords on their own, Anthropic responded by forcibly logging affected accounts out and stripping stored payment data from those accounts. The goal was to cut off unauthorized access immediately and prevent any charges that attackers might generate through API usage or subscription features. SecurityWeek reported the story, and Anthropic has not publicly disclosed how many accounts were involved or how the company identified the compromised sessions.

The forced logouts and payment data removals represent a proactive step that is relatively uncommon among AI platform providers. Most credential-stuffing or account-takeover disclosures ask users to reset passwords themselves. Anthropic's approach shifts the burden back to the platform, at the cost of disrupting legitimate users who now need to log back in and re-enter payment information.

Infostealer malware is a category of credential-harvesting software that runs locally on a victim's machine. Once installed — typically through a phishing email, a malicious download, or a trojanized software package — an infostealer scans the device for saved passwords, session cookies, browser-stored tokens, and autofill data. It then exfiltrates that data to an attacker-controlled server. Because modern browsers store authentication cookies that can keep a user logged in across sessions, a stolen cookie can grant account access without ever knowing the underlying password.

Why it matters

In Anthropic's case, the stolen material was apparently sufficient to authenticate as Claude users. Attackers with access to a valid session token can use Claude's features — including API calls — under the victim's account and billing arrangement. By deleting payment data and terminating sessions server-side, Anthropic invalidated the stolen tokens and removed the financial incentive for continued access, since any further API usage could not be billed to a stored card.

Infostealer campaigns increasingly target accounts on AI platforms because those accounts carry real monetary value. A Claude Pro subscription or an active API key with a billing limit represents a ready-made compute budget that attackers can resell or use to run their own workloads at a victim's expense. The same dynamic has appeared with OpenAI and other platforms, and security researchers have noted a growing market for stolen AI credentials on underground forums.

Who is affected

The incident also surfaces a structural risk that goes beyond password hygiene. Even users with strong, unique passwords are vulnerable if malware is running on their device, because the threat operates at the session layer rather than the credential layer. This means the standard advice — use a password manager and enable two-factor authentication — is necessary but not sufficient when an active session cookie can be lifted directly from a browser process.

Any Claude user who has experienced a forced logout should assume their device may be infected with infostealer malware. The immediate symptom — being logged out unexpectedly and finding payment information removed — is the signal Anthropic has sent to those accounts. Developers who use personal machines to manage Anthropic API keys face additional exposure, because a compromised device could also leak API keys stored in environment files, shell profiles, or local configuration directories, not just browser sessions.

Organizations that have employees accessing Claude or building on the Anthropic API on shared or unmanaged devices carry elevated risk. Enterprise security teams should treat an Anthropic-initiated forced logout as an incident indicator and begin a wider investigation into what other credentials may have been exfiltrated from the same machine. Infostealer infections are rarely limited to a single target service.

What to watch next

Users who have been logged out should not simply log back in and re-enter payment data before verifying their device is clean. Running a reputable malware scan and checking for recently installed unfamiliar software are minimum steps. Rotating any API keys associated with the affected account is equally important, since those are not automatically invalidated by a session termination. Anthropic has not confirmed whether API keys are being rotated on its end as part of the same protective action.

Watch for Anthropic to release a formal incident notice with a clearer account of scope and the timeline of the campaign. If the company has not already done so, it would be worth confirming whether it plans to notify affected users directly beyond the forced logout, and whether additional controls such as device trust or session anomaly detection are being added to Claude's authentication layer.