Skip to content
HomeNewsConceptsGuidesToolbox
AboutSubscribeUA
Subscribe

AI Today Brief

The daily AI-engineering brief. Built in public. EN · UA.

XTelegramLinkedInYouTubeRSS

Follow AI Today Brief on LinkedIn for daily AI-engineering updates and the weekly “5 shifts that changed how developers work” PDF.

Explore

NewsDigestsConceptsGuides

Company

SubscribeAdvertiseAbout

Legal

Editorial policyAI disclosurePrivacyTerms

© 2026 AI Today Brief. All rights reserved.

  1. Home/
  2. News/
  3. Token & cost optimization/
  4. Infostealers Hijack Active Claude Sessions to Drain Paid Subscriptions
Token & cost optimization

Infostealers Hijack Active Claude Sessions to Drain Paid Subscriptions

Anthropic disclosed that multiple infostealer malware families are harvesting authenticated browser cookies to hijack Claude accounts across Windows and macOS. Because attackers reuse active session tokens, they bypass two-factor authentication and passwords to consume expensive API usage and subscription compute.

September 1, 2026· 4 min read
OKCurated by Oleksandr Kuzmenko, AI Product Engineer·Updated September 1, 2026·Sources cited on every story
AI-assisted · editor-reviewed·How we use AI
Infostealers Hijack Active Claude Sessions to Drain Paid Subscriptions

Impact: High

Why it matters

You need to audit active sessions, clear compromised browser tokens, and consider hardware security keys to prevent unauthorized spend on your AI developer accounts.

TL;DR

  • 01Six infostealer families are actively exfiltrating Claude session tokens on Windows and macOS.
  • 02Cookie hijacking completely bypasses traditional passwords, SSO, and 2FA prompts.
  • 03Anthropic has revoked affected sessions and refunded unauthorized compute charges.

Key facts

Identified Malware Families
6 (Vidar, LummaC2, StealC, RedLine, Acreed, Atomic Stealer)
Targeted Operating Systems
Windows and macOS
FBI Initial Access Attribution
>70% in ransomware campaigns

The Mechanism of Cookie Hijacking

Anthropic's security team identified active credential theft campaigns harvesting session cookies from developer workstations. The identified malware strains include:

  • Windows: Vidar, LummaC2, StealC, RedLine, and Acreed.
  • macOS: Atomic Stealer.

Because these infostealers capture active session tokens rather than raw passwords, they bypass both standard password authentication and hardware-independent 2FA/SSO checks. The hijacked sessions allow attackers to access paid Claude Pro features and enterprise API quotas without triggering traditional login alerts.

Mitigation and Account Hardening

To remediate compromised developer environments:

1. Revoke Sessions: Navigate to Claude account settings and force-logout all active devices. 2. Scan Host Systems: Remove persistent local malware before entering fresh credentials to prevent immediate re-theft. 3. Hardware Keys: Migrate authentication policies to WebAuthn hardware security keys (FIDO2) which cannot be exfiltrated via simple cookie theft.

✓ When to use

  • Auditing developer workstation security policies across teams with shared Claude API keys or seats.
  • Investigating unexpected spikes in monthly Claude Pro usage or sudden token limit exhaustion.

What to do today

  • →Log into Claude settings and revoke all active sessions across web and desktop clients.
  • →Scan development machines for LummaC2, Atomic Stealer, and related credential-theft binaries.
  • →Implement WebAuthn hardware security keys for critical AI workspace administrative accounts.
#Claude#Anthropic#Claude Pro

Sources

  • Pulse of Nations: Infostealers Hijack Claude AI Sessions
ShareShare on XShare on LinkedIn
← Previous storyUnitpost Releases Open-Source Inbox-Safe Email Components and Agent SkillNext story →Google Releases TimesFM-3 for Zero-Shot Multivariate Time-Series Forecasting

Related stories

  • Token & cost optimizationShared Claude Chats Indexed by Search Engines Due to Missing Noindex Headers
  • Token & cost optimizationGLM-5.3-Flash Slashes Agentic Inference Costs by up to 50x
  • Token & cost optimizationOpenAI Codex Sol Reasoning Model Burns 5-Hour Limit in Minutes
  • Token & cost optimizationQuantization-Aware Healing Recovers 4-Bit LLMs Beyond Full-Precision Performance

Email digest

Get the morning AI brief

One email a day — the stories that matter for engineers, founders and tech leads. Human-edited, with links to primary sources.

  • ✓120+ sources scanned daily
  • ✓Edited by a human
  • ✓1 email per day
  • ✓EN + UA

By subscribing you agree to the privacy policy.