What happened
A report by Anthropic, a company that develops and monitors AI models, revealed that Iran used an advanced AI model developed in the US to track American warships in the Middle East. The model, called Claude, was used to gather information from ship transponders, military photos, and commercial satellite images to identify vulnerabilities in the ships' communication systems.
The report also stated that Iran used Claude to build a system for collecting and analyzing information on hundreds of Israelis and Jews in the diaspora, including individuals connected to government and non-governmental organizations in Israel. Additionally, Iran used Claude to build campaigns, target databases, and fake social media accounts, and during the war with Israel, attributed false claims to Western research institutions to make them appear credible.
Anthropic also reported that it had blocked the activity of an Israeli-Singaporean intelligence company, S2T Unlocking Cyberspace, which had developed a system for monitoring and analyzing social media users in Iran and Gulf countries based on location, demographic characteristics, and political views.
The report highlighted the potential dangers of AI models being used for malicious purposes, including the development of new weapons, aggressive disinformation campaigns, espionage, and even the development of biological weapons. Anthropic noted that it had identified five cases of biological research using Claude in recent months, and that this was the first time a private AI company had published findings on the potential use of its platform for developing biological weapons.
