🚨🇮🇷🇺🇸 An Iran-linked group used Anthropic to find U.S. warships and recommend TARGETS
Anthropic's report says the actor fed Claude publicly available material, satellite imagery, ship and aircraft tracking, even captions on military photos the Pentagon posted itself, and got back a picture of American naval movements with targeting recommendations.
The account was banned and the government notified.
Three Iranian state-aligned accounts were also caught running influence campaigns, inventing quotes from Brookings and RAND to make propaganda look credible.
Claude is blocked in Iran. A VPN and a foreign phone number get around that.
Georgetown's Sam Bresnick puts the problem plainly: it is getting much harder for the U.S. military to hide what it is doing.
Nikita Shah at CSIS asks what anyone expected when models this capable were made this accessible.
The countermeasure under review is the part to watch.
CBS saw a proposal to build fake AI personas that mimic real life patterns and muddy the military's digital trail.
Washington spent this war learning that its own public data, ad IDs, ship trackers, unit photos, is an intelligence feed for Tehran.
Fixing that means an American military that lies about itself online by design, and nobody has thought through what that does to a free press or a public that already distrusts what the Pentagon says.
Source: CBS News, Anthropic / Writer: Daniel
💬 18🔁 25
♥ 144👁 55K













