Influence operations · GTG-54004

Coordinated inauthentic behavior in Kenya: GTG-54004 case

Source notice

This page is an English translation of the relevant content from Anthropic's September 2026 report, pages 75–78. Any "taking sides" with a political figure is the report's description of the content's orientation, not this site's political judgment.

According to the report, one operator used Claude to generate batches of exactly 50 Kenyan political posts, with explicit instructions to make them look like spontaneous grassroots comments. The report found no evidence of government involvement and assessed that the operation stayed within its fake-account network without reaching real users. Once again, output volume is not the same as influence.

What happened

We identified and removed an account used by a single actor to mass-produce Kenyan political content. The operation was constructed to look like spontaneous grassroots public sentiment. Across multiple sessions, the actor used Claude to generate batches of exactly 50 tweets, with explicit instructions to make the posts look like spontaneous grassroots comments rather than a coordinated campaign.

The network focused its messaging on key Kenyan political issues and figures. Most of the AI-generated tweets praised Energy Cabinet Secretary Opiyo Wandayi for blocking a planned electricity tariff hike by Kenya Power, using the hashtags #PowerReliefKE and #PoweringTheNewKenya to boost visibility. In addition, the network pushed reports claiming that Kenya's United Opposition Alliance was splitting ahead of the 2027 general election, directly targeting politician Rigathi Gachagua and former President Uhuru Kenyatta.

Separately, the actor ran the same AI workflow for Kenyan retail brands under the marketing persona 'SHANKI'/'Elkins Marketer'. We found no evidence of government involvement, and the operation appears consistent with a fully domestic Kenyan effort.

Our investigation revealed a highly structured political astroturfing effort pushing the same unified message about the electricity tariff hike across different conversations. We assessed the network's impact using the Breakout Scale and classified it as Tier 1. The operation was fully isolated within the fake-account network and local influencers on a single platform, and failed to reach or influence any real users.

Although we do not know the exact real identity of the people behind the operation, we believe this is a local Kenyan political astroturfing campaign. The network's pro-government tone and specific hashtags suggest it may be aligned with the ruling coalition, although we have not confirmed the exact organization responsible.

What the AI did

The AI's role was to generate posts in fixed batches and with a unified message that were designed not to look unified. This is precisely the classic mechanism of 'coordinated inauthentic behavior': the content is consistent, while on the surface it looks like many people speaking individually.

What the report observed

The report assessed it as Tier 1 on the Breakout Scale: the operation was fully isolated within a single-platform network of fake accounts and local influencers, and failed to reach or influence real users.

Therefore, even though the content took a side on a particular administrative measure, it cannot be said to have changed public opinion.

Schematic: batch-generated posts staying within the fake-account network.
Original report figure: schematic related to the coordinated posting network. The text in the image is the original English.

Confirmed & unknown

Confirmed

  • The report confirms generation in batches of 50 with instructions to disguise as spontaneous comments
  • The report confirms no evidence of government involvement was found
  • The report assesses the operation did not break out of the fake-account network or reach real users

Unknown

  • The operator's real identity — the report states it is unknown
  • Whether the same person operated on other platforms — the report has not ruled this out entirely
  • Whether the marketing persona and the political posts served the same client — the report has not confirmed this

Platform response

Anthropic removed the account.

Limits of response:Removing the account does not mean all posts already published on social platforms have disappeared.

Takeaways

  • A large number of 'ordinary people' comments with consistent tone and hashtags appearing in a short time may be coordinated production.
  • An operation that does not break out of its own network shows no evidence of having swayed an election.
  • Domestic commercial marketing and domestic political disguise can share the same generation workflow.

Sources