Iran used AI to develop military, propaganda strategies, Anthropic report reveals

Iran repeatedly used American-made artificial intelligence systems to target US naval forces in the region, target Israeli and Jewish diaspora organizations, and develop phishing strategies inside the country, a Thursday Anthropic report revealed. 

Anthropic described what it called an “Iran-nexus threat actor” that used Claude, the company’s flagship product, to develop targeting recommendations against US Navy targets. 

The actor, according to Anthropic, used AI to identify and track naval positions and compile information, including “a roster of US personnel scraped from captions on public military photographs; publicly accessible ship and aircraft transponder identifiers; commercial satellite-imagery query scripts; and an inventory of public websites that exposed US naval movements.”

The unnamed organization also used AI to research vulnerabilities in shipboard systems, including “maritime VSAT terminals, Cisco communications equipment, and industrial control products,” according to the report.

This was one of the many findings revealed in the AI giant’s 154-page report, which delineated the dangers of US and Western adversaries such as Iran, China, and Russia utilizing the technology to pose security threats.

The report comes as debate over the dangers rages among policymakers and AI company employees, even as Anthropic refused the Pentagon’s February request to remove safety guardrails regarding mass domestic surveillance and fully autonomous weapons.

A Friday Wall Street Journal report revealed that a top researcher at Anthropic quit last week, citing fears that AI companies were developing uncontrollable systems. 

'AI could kill all humans,' top Anthropic scientist says

Evan Hubinger, a top scientist at Anthropic, wrote in a social-media post, “We really do earnestly believe AI could kill all humans!”

“I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to,” the WSJ cited him as saying. 

Anthropic’s report revealed that all groups showed similar methodology. “They ran multi-victim campaigns using agentic AI that would previously have required teams of operators. They built custom tools, executed intrusions, and processed stolen data at volumes no individual human operator could manage manually,” the company explained.

In addition to targeting US military assets, Iran also developed an “automated, open-source intelligence identity-profiling harness” that targeted hundreds of Israeli governmental and non-governmental individuals, as well as Jewish diaspora organizations.

The actor used the system to collect and analyze open-source data, developing a pipeline that made malware harder to detect.

Iran using AI to survey, profile Iranians

Another two “Iranian nexus actors” used the company’s model to surveil and profile 6,388 Iranians in one year. The AI company claimed that Iran built an “analyst and production studio” used to run a social network analysis of over 155,216 tweets.

Among other domestic surveillance strategies were automatic license-plate recognition, mobile-device identifier interception, and private Telegram data analysis.

These operations targeted Iranian minorities, dissidents, and opponents of the Iranian regime abroad, according to the report.

Actors also used AI to run campaigns tied to “Jihad al-Tabyin,” or explanatory jihad, to produce propaganda in what they called a “soft war” or “cognitive warfare” program: a non-military plan to shape public opinion at home and abroad, the report explained.

Several Iranian organizations, including the Islamic Culture and Communications Organization (ICCO), were using AI to distribute Islamic Revolutionary Guard Corps (IRGC) - aligned narratives, said the report.

The organizations used the technology to build operational manuals, create persona systems, develop content strategies, and transform official government material into targeted messaging.

Yemeni group developing missiles with AI

Anthropic also claimed that a northern-Yemeni group was using Claude to develop guidance software for high-end weapons, including a multi-stage ballistic missile.

This comes as the Iranian-backed Houthis consolidate control over the Bab al-Mandeb Strait. Anthropic said that its safeguards had blocked many of the group’s requests to develop the weapons, “but not all of them.”

The company confirmed that it had taken the necessary steps to ban the account, develop detections to prevent future use, and share intelligence with government authorities.

“We hope that the findings in this report will help other developers recognize similar patterns on their own platforms, give governments and civil society a clearer view of how emerging threats take shape, and strengthen collective defenses,” it said.