Anthropic: des infostealers pillent des sessions Claude actives
TL;DR
- Anthropic named six specific infostealer families by platform: Vidar, LummaC2, StealC, RedLine, and Acreed on Windows; Atomic Stealer on macOS, all commodity tools with no Claude-specific code.
- The attack bypasses 2FA entirely: session cookies carry an already-authenticated state, making password and MFA defenses irrelevant once a machine is compromised.
- Attackers likely harvested Claude sessions incidentally during bulk cookie sweeps, then selectively monetized AI credentials, suggesting organized resale infrastructure.
Anthropic a averti une partie des utilisateurs de Claude que des malwares infostealers courants siphonnaient leurs sessions authentifiées pour se connecter à leur compte et brûler leurs crédits d'usage, sans jamais présenter mot de passe ni code à deux facteurs. L'alerte est rapportée par BleepingComputer.
« We have recently become aware of a bad actor that is using common infostealer malware to steal Claude login sessions from people's computers », écrit la société. Sur les postes compromis, les « usage limits looked like they refilled and then drained while you weren't using Claude ». Anthropic pointe cinq familles sur Windows — Vidar, LummaC2, StealC, RedLine, Acreed — et Atomic Stealer (AMOS) sur un petit nombre de Mac.
Le vecteur n'a rien de spécifique au produit. Ces outils « typically arrives through downloads or malicious apps and steals information stored locally, including browser passwords, login cookies, and credentials belonging to other apps ». Un utilisateur affecté a publiquement raconté s'être infecté après avoir téléchargé un jeu piraté.
En réponse, Anthropic a déconnecté les comptes concernés, retiré les moyens de paiement enregistrés et remboursé les débits qu'elle qualifie de non autorisés, en tenant à préciser: « We have no reason to believe that this malware is related to Claude. » L'entreprise avertit que la déconnexion, seule, ne suffit pas. « Signing you out of Claude stops the stolen sessions, but it doesn't remove the malware. If it's still on your computer, your next login session could be stolen the same way. »
L'épisode s'inscrit dans un flux nourri de dossiers sur les risques cyber liés à l'IA suivis chez AI Weekly ces trois derniers mois.
Ce qu'en disent les autres médias
-
Security Affairs Lire →
Reproduces Anthropic's direct user notification language and adds the desktop-only attack surface detail; frames quota drain as a secondary signal of broader credential theft.
Your Claude session was likely one of the many things it collected. It appears that a bad actor has now started picking the Claude sessions out of what it collected.
-
Help Net Security Lire →
Details Anthropic's remediation protocol: forced session lockout, payment method removal, and refunds, plus a warning that phishing emails are now impersonating Anthropic's incident notices.
The malware identified in this campaign so far include Vidar, Lumma (LummaC2), StealC, RedLine and Acreed on Windows, and Atomic Stealer (AMOS)
-
eSecurity Planet Lire →
Contextualizes AI subscriptions as commodity cybercrime targets on par with banking credentials and provides step-by-step user remediation guidance.
We have recently become aware of a bad actor that is using common infostealer malware to steal Claude login sessions from people's computers.
-
The Cyber Express Lire →
Adds organizational risk framing: Claude accounts store conversation history, documents, and API keys, creating an enterprise endpoint security gap beyond individual quota loss.
-
Crypto Briefing Lire →
Flags LummaC2 and RedLine as established crypto wallet harvesters and notes Chrome Device Bound Session Credentials as an emerging architectural mitigation.
The stolen cookies gave attackers a backdoor into user accounts, letting them burn through paid usage quotas
Article original publié par bleepingcomputer.com
Lire l'article original →Titre original : Anthropic alerte que des malwares infostealers détournent les sessions Claude pour épuiser les quotas