What Anthropic Reported About the ‘Deadshot’ UAE-Linked Influence Operation

Anthropic’s September 2026 threat-intelligence report described an alleged covert influence operation, known internally as “Deadshot,” that it linked with high confidence to UAE government officials. According to Anthropic and reporting by France 24, the operation was primarily aimed at the Muslim Brotherhood while also seeking to counter criticism of the UAE’s alleged role in Sudan’s war. 11327

The key limitation is as important as the allegation: Anthropic documented attempted AI-assisted influence activity and disrupted related accounts, but it did not establish that the campaign’s materials reached their intended audiences, affected policymakers, or produced a measurable real-world outcome. 1

The reported objective: target the Muslim Brotherhood and deflect scrutiny over Sudan

Anthropic assessed that the operation’s strategic aim included a global effort to dismantle the Muslim Brotherhood. Its activities also focused on United Nations experts and other figures who had raised concerns about Abu Dhabi’s alleged involvement in Sudan’s conflict. 113

The Sudan connection matters because the campaign unfolded against allegations that the UAE supported Sudan’s Rapid Support Forces during the country’s civil war. The UAE has denied supporting the group. 1

Taken together, the reported materials appear to have served two connected purposes: promote anti–Muslim Brotherhood narratives and push back against accountability efforts concerning Sudan. That describes the operation’s apparent intent, rather than a finding that those efforts succeeded. 1

How Claude was reportedly used

Anthropic said the operators used its Claude AI system to generate a range of content related to Sudan, including:

• reports and advocacy documents;
• social-media material;
• content presented as UN testimony; and
• speeches intended for the 62nd session of the UN Human Rights Council. 113

France 24 reported that the proposed UN speeches were written to avoid mentioning the UAE. 1

This is a notable example of generative AI’s role in influence work: the reported use was not limited to public posts, but extended to research-style documents, advocacy language, and material intended for international-policy settings. Anthropic’s wider September report covered operations it disrupted between December 2025 and August 2026 across areas including influence operations, surveillance, cyber activity, scams, biological misuse, weapons development, and model distillation. 27

Fake accounts, impersonation, and targeting of critics

The Deadshot operation reportedly operated roughly 300 inauthentic accounts and impersonated a genuine Swiss human-rights organization. 1

Anthropic also said the operators compiled “counter-accountability dossiers” on UN Special Rapporteurs who had expressed concerns about the UAE and Sudan. The campaign reportedly profiled 18 members of the European Parliament and journalists as well. 1

These details suggest a campaign designed not only to distribute narratives but also to map, discredit, or influence people involved in scrutiny of the UAE’s conduct. Still, the available reporting does not independently verify every operational detail, and the underlying attribution remains Anthropic’s assessment. 1

What Anthropic did—and what remains unknown

Anthropic said it disrupted the operation by banning associated accounts. Its threat-intelligence program says it identifies and disrupts attempts to use Claude for malicious activity, using the cases to improve defenses and inform wider security efforts. 2227

However, the report could not determine whether the generated materials were delivered to intended recipients, submitted in the venues described, or influential in policy decisions. 1

That distinction prevents a broader conclusion: evidence that an AI system was used to prepare influence material is not, by itself, evidence that the material changed minds or policy. The Deadshot case is therefore best understood as a reported and disrupted attempt at AI-enabled influence, with claimed links to UAE officials, rather than proof of a successful campaign.

UAE response

France 24 reported that the UAE had not responded to the allegations at the time of its publication. 1