Anthropic's Claude has become so popular that criminals are now using its name as a Trojan horse to distribute sophisticated malware. Security researchers at Morphisec discovered that RevStealer, a new credential-stealing malware, is spreading through fake desktop versions of Claude Opus 5 posted on GitHub, exploiting users' trust in the AI company's brand to compromise their most sensitive data.

What Is RevStealer and How Does It Work?

RevStealer is a trojanized Electron application, meaning it masquerades as a legitimate desktop program while secretly running malicious code. The malware was distributed through a GitHub repository called "claude5opus/Claude-Opus-5-Free-Desktop," which offered a roughly 101 megabyte archive named ClaudeOpus5-desktop.zip. When users downloaded and ran this fake application, they unknowingly installed malware designed to harvest their most valuable digital assets.

The malware's infection chain begins silently. Once executed, the Electron loader checks whether the victim's computer has sufficient memory, CPU cores, and graphics hardware. It also screens for telltale signs of malware-analysis environments, such as specific usernames and hostnames that security researchers use to study threats. If the system passes these checks, RevStealer attempts to add the Windows AppData folder to Microsoft Defender's exclusion list using hidden PowerShell commands, effectively blinding the antivirus software to its presence.

What Data Does RevStealer Actually Steal?

RevStealer's appetite for data is comprehensive and alarming. The malware targets a wide range of sensitive information that could devastate victims financially and personally:

• Browser Data: Credentials and session information from Chromium and Firefox browsers, giving attackers access to logged-in accounts across websites.
• Password Managers: Encrypted vault files from Bitwarden, 1Password, KeePass, and LastPass, potentially unlocking all stored passwords if cracked offline.
• Cryptocurrency Wallets: Private keys and wallet data from more than 50 different cryptocurrency platforms, enabling direct theft of digital assets.
• Windows Credentials: Passwords stored in Windows Credential Manager, which often contain access to corporate systems and cloud services.
• VPN and Remote Access Tools: Configuration files that could allow attackers to impersonate victims on corporate networks.
• Personal Files: Screenshots, clipboard contents, and selected documents that may contain sensitive information, financial records, or personal communications.

Notably, some stolen databases and encrypted vault files are collected raw without being decrypted locally, suggesting the attackers plan to process and crack them on their own servers away from the infected system.

How Does RevStealer Communicate and Avoid Detection?

RevStealer uses multiple layers of obfuscation to hide its tracks. The malware employs concealed API resolution, encrypted runtime strings, and indirect Windows system calls, all designed to make it harder for security tools to detect its activities. It also performs additional virtual machine and sandbox checks to avoid triggering alerts in security research environments.

The malware's command-and-control (C2) infrastructure, which is the server that receives stolen data and sends commands back to infected machines, uses meta7[.]archscreen68[.]one:443 as its primary address. However, RevStealer includes a clever failover mechanism: if that server becomes unavailable, the malware queries a Polygon smart contract through public blockchain RPC providers to retrieve a replacement server address. This approach allows the attackers to change their infrastructure without rebuilding the malware, making it significantly harder for law enforcement or security companies to shut down the operation.

How to Protect Yourself From Claude Impersonation Malware

Users and defenders can take several concrete steps to avoid falling victim to this threat and to detect infections if they occur:

• Avoid Unofficial Downloads: Never download unofficial "free" versions of paid AI software, particularly executable files from unverified repositories. Anthropic's Claude is available through official channels only; any desktop version claiming to be free from GitHub or other third-party sources is almost certainly malicious.
• Monitor Defender Exclusions: Defenders should actively monitor for unexpected additions to Microsoft Defender's exclusion list, especially any that cover the AppData folder, which is a common staging area for malware.
• Watch for Known Indicators: Security teams should monitor for the reported mutex identifier Global\5B908BC4, the staging path %USERPROFILE%\AppData.9qdz1mlg, and the C2 server address meta7[.]archscreen68[.]one:443.
• Treat Infections Seriously: Any system confirmed to be infected should be treated as fully compromised. All passwords, browser sessions, and relevant access tokens should be revoked or rotated immediately, and the device should be wiped and reinstalled.

Why Is Anthropic's Brand Being Targeted?

The use of Claude's branding is not accidental. Anthropic has become one of the most trusted names in artificial intelligence, with Claude gaining significant popularity among developers and everyday users. By impersonating Claude, attackers exploit this trust to lower users' guard. The malware campaign also used game-cheat-themed websites as alternative distribution channels, suggesting the attackers are casting a wide net across multiple communities.

Importantly, there is no indication that Anthropic itself was compromised or that the company's systems were breached. GitHub served only as one distribution channel identified in the campaign. The attack is a case of brand impersonation rather than a breach of Anthropic's infrastructure. However, the incident underscores a growing risk: as AI companies become more prominent, their brands become more valuable targets for social engineering and malware distribution.

The RevStealer campaign demonstrates a troubling trend in cybercrime: attackers are increasingly leveraging the reputation and trust of legitimate technology companies to distribute malware. Users should remain vigilant about downloading software only from official sources and should be especially cautious with any claims of "free" versions of premium AI tools. For security teams, the incident serves as a reminder to monitor for suspicious system changes, particularly modifications to antivirus exclusion lists, which are a classic sign of malware attempting to hide its presence.