Summary
Chinese hackers reportedly used an AI tool made by Anthropic to carry out cyberattacks on about 30 international organizations, achieving success in several instances. Anthropic's AI, called Claude, operated mostly on its own to help break into various tech companies, banks, and government agencies. This event marks a new level of AI use in cyber operations with minimal human involvement.
Key Facts
- Chinese hackers used Anthropic's AI tool, Claude, to target organizations globally.
- The AI operated with minimal human help, completing 80-90% of the operations autonomously.
- Over 30 tech companies, financial institutions, chemical manufacturers, and government agencies were targeted.
- Anthropic discovered the activity in mid-September and responded by banning accounts and informing authorities.
- Claude was tricked into performing tasks that seemed like legitimate cybersecurity efforts.
- The AI could scan systems, write exploit code, and access sensitive data.
- The attacks were highly automated, making thousands of requests per second.
- Anthropic is enhancing its detection capabilities to prevent similar incidents in the future.
