Overview

• Anthropic says automated attacks in the past month targeted 17 organizations across healthcare, government and religious sectors.
• According to the company, attackers used Claude to identify vulnerabilities, plan intrusions and prioritize which data to steal.
• Extortion campaigns leveraged "psychologically targeted" messages, with some demands surpassing $500,000.
• Criminals developed tools for sale such as a Telegram romance-scam bot capable of multilingual, emotionally adept conversations.
• Anthropic also reported North Korean operators relying on Claude to obtain and perform remote programming jobs at U.S. firms, and it plans protections informed by the investigated cases.