All corrections
X September 11, 2026 at 09:21 AM

x.com/ComfortablySmug/status/2098147407666049324

2 corrections found

1
Claim
automate dossiers on Americans
Correction

Anthropic’s report does describe dossier-building, but not on Americans. Its Iran-linked cases involved profiling Iranians and building psychographic dossiers on people inside Iran.

Full reasoning

Anthropic’s September 2026 report includes dossier- and profiling-related activity, but the targets it names are not Americans.

For the Iran-linked influence operation tied to MEK/NCRI, Anthropic says the network scraped hundreds of social-media channels to build detailed profiles of individuals inside Iran, then built psychographic dossiers on dozens of specific individuals in Iran. In Anthropic’s separate Iran-linked surveillance case, the company says the operators maintained records on Iranian nationals, surveilled and profiled 6,388 Iranians in a year, and named 39 Iranian opposition and diaspora accounts.

So the post changes the report’s stated targets. Anthropic’s own write-up points to dossiers/profiles on Iranians or people inside Iran, not “Americans.”

2 sources
  • Detecting and countering misuse of AI: September 2026

    “The operation successfully scraped over 500 social media channels to build detailed profiles of individuals inside Iran… The network analyzed roughly 51,944 archived messages… to build detailed psychographic dossiers on dozens of specific individuals in Iran.”

  • Detecting and countering misuse of AI: September 2026

    “We identified a seven-department organization with offices across Iran’s provinces that claimed to maintain an identity-record database of Iranian nationals, and to surveil and profile 6,388 Iranians in a single year… [It] ran a social-network-analysis pipeline that named 39 Iranian opposition and diaspora accounts.”

2
Claim
Iran using Claude to target Navy bases
Correction

Anthropic’s Navy-related case was not Iranian. Its report says the Navy case involved a China-based actor benchmarking against U.S. Navy systems for a proposal to the People’s Liberation Army Navy.

Full reasoning

Anthropic’s September 2026 threat report does describe a Navy-related misuse case, but it attributes that case to a China-based actor, not Iran. In the conventional-weapons section, Anthropic says a China-based actor used Claude to draft anti-torpedo fire-control documents and to generate a briefing on U.S. Navy systems, and that the actor was associated with a Chinese defense manufacturer aiming to support the People’s Liberation Army Navy.

By contrast, Anthropic’s Iran-linked cases in the same report are described as influence operations and domestic surveillance/profiling of Iranians. The Iran section says those actors used Claude to build campaign plans, doctrine manuals, persona systems, target databases, and ministerial planning documents for propaganda efforts. That means the post is conflating separate case studies and incorrectly assigning the Navy-related one to Iran.

2 sources
  • Detecting and countering misuse of AI: September 2026

    “GTG-17001: Disrupting a China-based operation… *Third, the actor used Claude to benchmark their own system against specific US anti-torpedo and anti-submarine programs… and generated a Chinese-language briefing on US Navy systems…* We assess the actor was associated with a Chinese defense industry manufacturer aiming to produce a weapons specification and acquisition proposal for the People’s Liberation Army Navy.”

  • Detecting and countering misuse of AI: September 2026

    “GTG-34001: Disrupting Iranian state-aligned influence operations… We identified and removed three Iranian state-aligned accounts… In each case, the operation relied on Claude to build campaign plans, doctrine manuals, persona systems, target databases, and ministerial planning documentation.”

Model: OPENAI_GPT_5 Prompt: v1.16.0