What happened
A recent report by Anthropic revealed that Iranian groups have been using the AI model Claude to wage war against American citizens and soldiers. The groups used Claude for various purposes, including gathering publicly available data to generate targeting recommendations, tracking naval ships, and identifying potential vulnerabilities in naval ships' electronics systems.
The Iranian groups also used Claude to build a Python script that extracted information from military photographs available online, creating a roster of possible American personnel in the area. Additionally, they used Claude as a research tool to find potential vulnerabilities in naval ships' electronics systems.
Anthropic banned the accounts involved in this activity, but not before the Iranian groups obtained a significant amount of information about the US military. The report also revealed that Iranian state-aligned influence operations used Claude to sway public opinion on the ongoing battle, targeting both Iranian people and American citizens.
The influence operations involved generating robust campaign strategies, guides, databases, and other documents, and attempting to make their propaganda efforts appear organic or grassroots. Anthropic was able to detect the abuse and trace the activity of the accounts, but the incident highlights the potential risks of AI technology being used against the US.
