A ransomware group called LunaLock targeted the digital art marketplace Artists&Clients in what security researchers are calling the first “AI extortion” attack. The group stole artwork from over 95,000 users and demanded a $50,000 ransom per account in Monero cryptocurrency.
What makes this attack different from typical ransomware is the threat: instead of just encrypting or leaking the data, LunaLock threatened to submit the stolen artwork into large-scale AI training pipelines. If the data gets fed into generative AI models, it becomes functionally impossible to remove, permanently embedding the artists’ intellectual property into AI systems.
How the Attack Happened
The breach started with spear-phishing emails sent to independent illustrators on the platform. The emails were disguised as urgent royalty notification invoices with trojanized attachments. Once opened, a custom loader ran on the victim’s machine.
The loader used Dynamic API Resolution with XOR-based string obfuscation to hide from static analysis tools. Critical function names like WriteProcessMemory and CreateRemoteThread were encoded in the data segment so that signature-based scanners couldn’t detect them.
The malware then injected a minified JavaScript module into the Windows Service Control Manager to disable Windows Defender’s real-time scanning. A hidden scheduled task named “SysUpdate” ensured the malware would restart on every system reboot.
Lateral Movement Through Stolen Tokens
Instead of cracking passwords or bypassing MFA, the attackers extracted active session tokens from Microsoft Teams and Slack clients. These tokens represent already-authenticated sessions, so the attackers moved through the organization’s systems with the authority of trusted employees.
They accessed cloud-hosted repositories and project management platforms without triggering any suspicious login alerts or MFA challenges. The stolen session tokens let them traverse the entire Artists&Clients infrastructure.
The AI Extortion Angle
Traditional ransomware threatens to delete or publish stolen data. LunaLock added a third option that’s specific to the AI era: feeding the artwork into AI training datasets. This creates a unique regulatory problem under GDPR’s “Right to be Forgotten,” since data that has been trained into an AI model cannot be functionally deleted.
Penalties under GDPR can reach up to 20 million euros or 4% of global annual turnover for severe breaches. The attack also exposed private client chat histories, commission archives, and creative portfolios across all 95,000 affected accounts.
Prevention and Indicators
Security teams should watch for files encrypted with the .lunalock extension, the hidden SysUpdate scheduled task, and unusual outbound requests to the group’s onion C2 domain. The malware also leaves evidence of JavaScript modules injected into the Windows Service Control Manager.
Researchers at the University of Chicago recommend using tools like Nightshade and Glaze to subtly alter high-value images before sharing online. These tools make artwork unusable for AI training even if the data is stolen, acting as a pre-emptive defense against AI laundering attacks.
There is currently no universal LunaLock decryptor for this specific AI-focused variant. Older versions had some decryption solutions, but the 2025 campaign uses different encryption methods.
Frequently Asked Questions
What is LunaLock ransomware?
LunaLock is a ransomware group that targeted digital artists through the Artists&Clients platform. They stole artwork from 95,000 users and threatened to submit it to AI training datasets if the ransom wasn’t paid.
How much was the LunaLock ransom demand?
The group demanded $50,000 in Monero per account. The total affected 95,000 user accounts on the Artists&Clients marketplace.
Can stolen artwork be removed from AI models?
No. Once data is trained into generative AI models, it cannot be functionally removed. This is why the LunaLock threat is considered particularly damaging, and why it creates GDPR compliance issues around the Right to be Forgotten.
How can artists protect themselves from AI extortion attacks?
Artists can use tools like Nightshade or Glaze to alter their images before sharing online. These tools make the images unusable for AI training. Artists should also be cautious of phishing emails disguised as royalty notifications or invoices.
Is there a decryptor for LunaLock ransomware?
No universal decryptor exists for the current AI-focused LunaLock variant. Older versions of the ransomware had some decryption solutions, but the 2025 campaign uses different encryption methods.
