The Morning Risk Report: How Hackers Steal Millions in AI Computing Power

A facts-only brief, written by a local AI model from this article in The Wall Street Journal, November 11, 2026. It may contain mistakes, and it is not the article: the full story is at wsj.com.

Hackers stole over $1.1 million in computing power from cloud provider Railway after beginning their probe on a Sunday night in mid-October. The attackers created thousands of accounts to access artificial intelligence models, a tactic researchers at Google’s cloud-security division, Wiz, identify as part of a broader trend of “LLM jacking.” These illicit operations often supply stolen compute to transfer stations that sell access to U.S. models in restricted countries like China, Russia, and Iran.

Background

• Wiz researchers note that some attacks target API keys to clone U.S. AI models using a technique called distillation.
• Google states that corporate accounts can suffer up to $100,000 in fraudulent charges daily, with total organizational losses reaching $1 million.
• Anthropic has warned that transfer stations are frequently used to facilitate the illegal cloning of American AI models.

Read the full article at wsj.com

More from Wednesday, November 11: