Anthropic, a US-based artificial intelligence development company, stated in its threat intelligence report that cybercriminals are increasingly using AI to carry out cyberattacks.

The company added that its chatbot “Claude” has been illegally used to breach networks, steal and analyze data, and craft “psychologically targeted extortion demands.”

In some cases, attackers threatened to release stolen information unless they received sums exceeding $500,000.

The company said that in the past month alone, 17 organizations in healthcare, government, and religious sectors were targeted.

“Claude” helped attackers identify security vulnerabilities, pinpoint target networks, and determine which data to extract.

Jacob Klein, Anthropic’s director, told the tech site “The Verge” that such operations previously required specialized expert teams, but AI now enables a single individual to launch sophisticated attacks.

Anthropic also documented cases of North Korean operatives using the “Claude” chatbot while impersonating remote programmers working for American companies to “fund North Korea’s weapons programs,” with AI assisting them in communicating with employers and performing tasks they lacked the skills to do themselves.

Historically, North Korean workers had to undergo years of training for such purposes, but “Claude and other models have effectively removed this barrier,” the company said.

Criminals also devised AI-powered scam schemes sold online, including a Telegram bot used in romance scams that emotionally manipulate victims in multiple languages to extort money.

Anthropic confirmed it has already implemented preventive measures to limit misuse, but attackers continue to seek ways to circumvent them.

The company said lessons learned from these incidents are used to enhance protection against AI-powered cybercrime.