Hacker admits using AI malware to breach Disney employee data

The intersection of AI tools and cybersecurity continues to evolve dangerously, as demonstrated by a recent case where malicious code embedded in an AI image generation tool led to a major data breach at Disney. This incident highlights how threat actors are exploiting the growing popularity of AI applications to distribute trojans that can compromise high-value corporate targets and personal information.

The big picture: A California man has pleaded guilty to hacking a Disney employee by distributing a malicious version of a popular open source AI image generation tool that stole sensitive corporate and personal data.

Key details: Ryan Mitchell Kramer, 25, admitted to publishing a fake AI art creation extension on GitHub that contained hidden malicious code giving him unauthorized access to users’ computers.

• The program, identified by researchers as “ComfyUI_LLMVISION,” masqueraded as an extension for the legitimate ComfyUI image generator but secretly copied passwords, payment card data, and other sensitive information.
• To better disguise the malicious code, Kramer used file names referencing reputable AI companies like OpenAI and Anthropic.

How the attack unfolded: After a Disney employee downloaded the malicious extension in April 2024, Kramer gained access to private Disney Slack channels and exfiltrated approximately 1.1 terabytes of confidential data.

• In July, Kramer contacted the employee pretending to be part of a hacktivist group and later released the stolen information publicly when he received no response.
• The leaked data included not only Disney’s private corporate information but also the employee’s personal banking, medical, and other sensitive details.

Why this matters: The case illustrates how threat actors are exploiting the enthusiasm around AI tools to distribute sophisticated trojans targeting high-value corporate environments.

The legal consequences: Kramer has pleaded guilty to unauthorized computer access and threatening to damage a protected computer, with his first court appearance expected within weeks.

• In his plea agreement, Kramer admitted to similarly compromising two additional victims who installed his malicious extension.
• The FBI is continuing its investigation into the matter.