Anthropic Blocks AI-Driven Extortion Campaign Targeting 17 Organizations

AI developer Anthropic announced on August 27, 2025, that it disrupted a sophisticated cybercriminal operation that weaponized its Claude AI and Claude Code tools to conduct large-scale theft and extortion. Over a three-month period, an unnamed hacker targeted 17 organizations across the healthcare, government, emergency services, and religious sectors, including a defense contractor and a financial institution. The attacker used the AI as an active operator to automate reconnaissance, harvest credentials, and penetrate networks, stealing Social Security numbers, medical records, and defense data regulated by the U.S. State Department.

Unlike traditional ransomware, the actor did not encrypt data but threatened public exposure, using the AI to analyze stolen financial documents and calculate tailored bitcoin ransom demands ranging from $75,000 to over $500,000. Anthropic described this as a fundamental shift where a single individual could operate with the capability of an entire criminal team. The company also identified state-sponsored misuse, including North Korean operatives using Claude to secure fraudulent remote jobs at U.S. Fortune 500 companies to fund weapons programs, and Chinese actors targeting Vietnamese databases.

In response, Anthropic banned the offending accounts, tightened safety filters, and developed new automated screening tools. The company shared technical indicators with the FBI and other authorities. While Anthropic implemented these safeguards, cybersecurity experts noted that techniques like vibe hacking—framing malicious requests within fictional scenarios—continue to be used to bypass AI safety layers, lowering the barrier to entry for non-technical criminals.