Photo: PatrickAssale / Shutterstock.com
Unknown threat actors have abused Anthropic’s Claude AI chatbot to launch an “influence-as-a-service” offering that interacts with legitimate accounts on Facebook and X. The campaign is assumed to be financially motivated and has created more than 100 unique personas engaging with thousands of legitimate accounts.
Once the attackers had their infrastructure ready, they used these accounts to promote clients’ political perspectives. Hacker News reports that these political perspectives either supported or undermined European, Iranian, Keynan, and UAE interests.
Claude was used not only for content generation but also to decide when these bot accounts would comment, like, or share posts from the legitimate accounts they were targeting. The overall activity is consistent with state-sponsored campaigns, but no affiliation has been discovered.
These sophisticated campaigns consistently tracked and updated persona attributes, engagement history, and narrative themes across multiple accounts. The bot accounts were also instructed to respond with humor or sarcasm when other accounts accused them of being bots. At least four different campaigns using the network have been found, and the operation has been taken down.
Anthropic’s report on the campaign includes other malicious activity as well. One threat actor used Claude to “enhance systems for identifying and processing exposed usernames and passwords associated with security cameras, while simultaneously collecting information on internet-facing targets to test these credentials against.”
Another abuse case was a recruitment fraud campaign that used Claude to enhance scam content targeting job seekers in Eastern European countries. Last but not least, Anthropic also caught an individual actor “with limited technical skills” developing a malware that would require much more expertise. The AI company wasn’t able to confirm successful deployment of any of these abuses.
In the News: Google fixes actively exploited Android flaw with May security update
