🚨 The Islamic Republic used AI to turn state propaganda into fake grassroots opinion across multiple languages and platforms. According to Anthropic’s latest threat-intelligence report, Iranian state-aligned operators were caught using Claude as a tool for a sophisticated influence operation designed to hide where Islamic Republic narratives were actually coming from. This went far beyond generating a few social media posts. Operators used AI to build fake personas, campaign strategies, targeting databases, amplification schedules and multilingual content in Farsi, Arabic, Urdu, Malay, Spanish and English. The objective was to make state messaging appear to come from ordinary citizens, independent commentators and foreign voices. Even more brazenly, they reportedly instructed Claude to write in the voice of an IRGC spokesman, then repackage state talking points to resemble independent analysis from major American institutions such as CSIS, Brookings and RAND. Those institutions are not implicated. Their credibility was effectively part of what the influence operation sought to imitate. A Mashhad-based operation developed a content factory called “Manjanegh” or “Catapult,” where activists and personas could take security-service material, remove the obvious IRGC fingerprint and distribute it across social media. The planned ecosystem stretched across X, Instagram, Telegram, TikTok, YouTube and Iranian platforms, combining personas, coordinated hashtags, paid amplification and carefully timed campaigns intended to manufacture the appearance of organic public opinion. And propaganda was only part of what Anthropic uncovered. Iran-linked operators also used Claude in work involving analysis of more than 155,000 tweets and the profiling of Iranians, while another effort involved a malicious Firefox extension disguised as a prayer-times tool. Other Iran-linked accounts used Claude for open-source research involving U.S. Navy ships and personnel, including ship tracking, satellite-imagery research and publicly available information on American personnel. State-aligned operators even used the AI for organizational planning surrounding the Supreme Leader’s funeral and succession. There is an important distinction: Anthropic did not document Claude autonomously operating some enormous army of thousands of bots. What it documented is arguably more calculated: a state influence apparatus using AI to industrialize propaganda, manufacture identities, translate narratives, disguise their origin and coordinate how they spread. Anthropic says the accounts were identified and shut down. The next time dozens of supposedly independent accounts suddenly repeat the same Islamic Republic narrative in different languages, remember that manufacturing the appearance of grassroots opinion was reportedly part of the strategy.