Iran just used our own top tech to wage war against us — what we learned is horrifying

If you think that only the U.S. military uses frontier AI models in combat, think again. Anthropic just caught several Iranian groups weaponizing Claude against American citizens and soldiers, and their multitiered campaign was aimed at targets in the region and within the United States.

According to a 154-page threat report released by Anthropic, Iran used Claude for a wide range of offenses aimed at the United States. In the section labeled “Military reconnaissance,” Iran clearly weaponized the platform to monitor and even target American soldiers during the ongoing confrontation in the Strait of Hormuz and surrounding areas.

The bad actors obtained a terrifying amount of information about the US military.

Anthropic described how an “Iran-nexus” used Claude against American naval personnel within the region in the following ways. They:

1. Gathered publicly available data and analyzed it to generate targeting recommendations and guides to be used against the U.S.
2. Created a system to help track naval ships, bolstered by commercial satellite imagery query scripts that extracted Earth image data and public transponder identifiers emitted from vessels.
3. Used Claude as a coding partner to build a Python script that dug through military photographs available online and created a roster of possible American personnel in the area.
4. Used Claude as a research tool to help the attackers find potential vulnerabilities in naval ships’ electronics systems.

Ultimately, Anthropic banned the accounts that were complicit in this activity. However, even though the threat was eliminated, the bad actors still obtained a terrifying amount of information about the U.S. military before they were discovered. Likewise, Anthropic also learned quite a bit about the Iranian operations through its platform, so the information goes both ways.

RELATED: America is losing the information war with Iran

Morteza Nikoubazl/NurPhoto/Getty Images

In another section titled, “Disrupting Iranian state-aligned influence operations on Claude,” three accounts with connections to the Iranian state engaged in influence operations — dubbed a “soft war” or “cognitive warfare” — meant to sway public opinion on the ongoing battle. The campaign targeted both the Iranian people, like those who took to the streets to protest the Iranian government earlier this year, and American citizens living in the United States. Anthropic found clear connections between the accounts and known Iranian propaganda groups, including:

• Islamic Culture and Communications Organization,
• Islamic Propaganda Office of Khorasan Razavi, and
• Islamic Propaganda Organization’s Bina Cultural Observatory.

Each group used Claude to generate robust campaign strategies, guides, databases, and other documents at a scale befitting a large team. They also focused on making their propaganda efforts appear like organic or grassroots campaigns to give off the perception of authenticity.

Anthropic noted that the groups partially attempted to hide their Iranian roots through VPNs and foreign phone numbers used to verify accounts when they were registered. However, users repeatedly mentioned identifying details in their prompts that gave away their origins, including their location, associated propaganda organization, and even the role they played within each group. Finally, Anthropic later traced some of the created materials to other online platforms that aligned with the Islamic Revolutionary Guard Corps, proving their connection.

This revelation brings several truths to light:

• American-made frontier AI platforms can and are being leveraged by foreign powers against American soldiers in live combat zones. While they may not have close system integration on par with our military’s GenAI.mil platform, these resources can still threaten real American lives on the battlefield.
• AI companies, like Anthropic, have the power to detect abuse within their platforms, but they may not always be quick to act. By the time Anthropic banned the Iranian accounts in question, they already produced at least some content to aid in their opposition against the United States.
• AI companies can trace the activity of every account registered on their platform. This is good news when punishing bad actors that want to exploit AI for evil. However, regular users are also subject to the same visibility and scrutiny.

As AI expands to every corner and facet of life, it was only a matter of time before our enemies tried to use American AI technology against us. This is why the U.S. military has taken steps to build its own AI platform stocked with an entire army of AI models. In this way, it can pick and choose the best attributes from each bot and assemble it all into a lethal war machine unlike anything else on the planet.

Like Blaze News? Bypass the censors, sign up for our newsletters, and get stories like this direct to your inbox. Sign up here!