Anthropic claims Chinese spies used chatbot for cyberattacks – Tech Digest

Share

Anthropic claims to have detected a sophisticated cyber espionage campaign orchestrated by Chinese government hackers using its Claude AI tool to perform automated attacks against nearly 30 organizations globally.

In a public blog post, Anthropic stated that hackers posed as legitimate cybersecurity workers and fed Claude a series of small, automated tasks. When chained together, these tasks formed what the company calls the “first reported AI-orchestrated cyber espionage campaign.”

Anthropic researchers expressed “high confidence” that the individuals behind the attacks were “a Chinese state-sponsored group.” The hackers reportedly used Claude’s coding assistance to build an unspecified program designed to “autonomously compromise a chosen target with little human involvement.”

Targets included major tech companies, financial institutions, chemical manufacturing firms, and government agencies. Anthropic asserts that the chatbot successfully breached various unnamed organizations, extracting sensitive data and sorting it for valuable information.

The company has since banned the hackers from Claude and notified affected companies and law enforcement.

However, the claim of a fully “AI-orchestrated” campaign faces scepticism from cyber experts, who question both the accuracy of Anthropic’s claim and its motives.

Critics argue that AI technology is still too unwieldy for truly autonomous, sophisticated cyberattacks. Anthropic itself admitted that Claude made mistakes during the process, such as generating fake login credentials and attempting to extract publicly available information.

“This remains an obstacle to fully autonomous cyberattacks,” Anthropic noted.

The announcement follows a similar report from OpenAI in February 2024, which detailed disrupting five state-affiliated actors, including some from China, who used their services for basic tasks such as querying open-source information and translating.

Anthropic argued that the solution to stopping AI attackers is to utilize AI defenders, claiming that the “very abilities that allow Claude to be used in these attacks also make it crucial for cyber defence.”

For latest tech stories go to TechDigest.tv