Anthropic official brand assets (anthropic.com)
A single operator ran roughly 300 fake social media accounts, a fabricated NGO, and ghostwritten UN testimony through Anthropic's AI assistant, the company says
Anthropic’s September 2026 threat intelligence report details one of the most elaborate state-linked influence operations ever conducted through a commercial AI system. A UAE-directed campaign, tracked internally as GTG-84002, used Claude to orchestrate what Anthropic describes as a “coordinated transatlantic and regional operation to dismantle the Muslim Brotherhood globally.”
The operation was run by a single actor the report identifies as “Deadshot,” who managed approximately 300 inauthentic social media accounts, created a fabricated NGO designed to impersonate a legitimate Swiss human rights organization, and produced ghostwritten testimony submitted to the UN Human Rights Council. Anthropic says it has high confidence the campaign traces back to UAE government officials.
What Deadshot actually built #
According to Anthropic’s report, published September 10, Deadshot used Claude to develop personal dossiers on 18 European Parliament members and journalists. Those dossiers appear designed to map potential pressure points and sympathies across European political institutions.
The campaign also generated counter-dossiers targeting UN special rapporteurs who had publicly criticized the UAE’s involvement in Sudan’s ongoing war.
By creating what appeared to be a Swiss human rights organization, the operator gave its messaging an air of neutral, institutional credibility. The ghostwritten UN Human Rights Council testimony was carefully scrubbed of any reference to the UAE, a design choice that Anthropic flagged as evidence of deliberate concealment rather than mere omission.
The UAE designated the Muslim Brotherhood a terrorist organization back in 2014, making the campaign’s objective consistent with long-standing Emirati foreign policy.
A pattern, not an anomaly #
Anthropic frames GTG-84002 as part of a broader wave of misuse it tracked between December 2025 and August 2026. The report catalogs multiple instances where Claude was leveraged in cyber operations and geopolitical influence campaigns across different regions and conflict zones.
OpenAI disclosed similar disruptions in 2024, identifying operations tied to Russia, China, Iran, and Israel. The Anthropic report stands out for the granularity of its attribution and the single-operator model it describes: one individual sustained 300 fake accounts, fabricated an entire NGO, drafted UN-ready testimony, and compiled intelligence dossiers on European lawmakers using a commercially available AI tool.
In response, Anthropic says it has banned all accounts associated with the campaign, implemented additional safeguards, and shared its intelligence with global partners. The company did not specify which governments or organizations received the briefing.
The UAE has faced sustained international criticism over its alleged role in Sudan’s civil war, which has displaced millions since fighting erupted in April 2023. The counter-dossiers targeting UN special rapporteurs who raised these criticisms suggest the influence campaign was not purely ideological — it was also aimed at blunting accountability mechanisms.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our