ANTHROPIC
Hackers are now automating entire attacks with AI
Anthropic says a Chinese state-backed hacking group used its AI model Claude to automate around 30 cyberattacks on companies and governments in September, as reported by the Wall Street Journal.
The company said 80%- 90% of the work was done by AI, with humans stepping in only to approve or correct parts of the process.
The hack fits a wider trend. Cyberattacks are becoming increasingly automated as hackers utilise AI to automate tasks.
Google recently reported that Russian groups are using LLMs to help generate malware commands.
The US has also warned for years that China uses AI to gather sensitive data, which China has denied.
Worth noting:
Most of the operation was automated through AI, needing very little human involvement.
Tech companies say multiple countries now use AI to support cyberattacks.
Anthropic confirmed data theft but did not disclose who was targeted.
Humans just press “Continue”
Anthropic believes this campaign was state-sponsored.
Four victims had data stolen, but their names were not shared.
None of the attempts on the US government succeeded.
Peak 2025 moment: even cybercrime is automated - MG