AI Astroturfing Disrupted
Anthropic disrupts a domestic Kenyan influence operation using Claude for astroturfing.
Anonymous
Key Points
1
Anthropic said the investigation began after a tip from OpenAI about repeat misuse of its platform; Anthropic then removed the account and the organisation behind the operation and developed detection measures for similar activity.
2
The actor also used an identical Claude-based workflow to promote Kenyan retail brands under the persona “SHANKI”/“Elkins Marketer.” The promotional material was repackaged as apparently organic posts, with links inserted into every fifth post.
3
Anthropic classified the operation as Category One on its Breakout Scale, the lowest tier, indicating that the campaign did not achieve a significant breakthrough beyond the originating network.
4
The operation was designated GTG-54004 and focused not only on opposition politics but also on the cost of electricity and perceptions of government performance.
5
The electricity-related messaging followed Kenya’s June 2026 withdrawal of Kenya Power’s tariff-review application. Energy Cabinet Secretary Opiyo Wandayi said existing tariffs would remain in force after consultations, while the Energy Ministry said the decision sought to shield households, businesses and industries from higher costs.
Anthropic said it disrupted a Kenyan influence operation where a single actor used Claude to mass-produce fake grassroots social-media posts supporting the government. Business Today Kenya reported the campaign generated 50-post batches promoting Energy Cabinet Secretary Opiyo Wandayi and attacking opposition figures like Rigathi Gachagua and Uhuru Kenyatta ahead of Kenya's 2027 election.
The actor used identical tactics to promote Kenyan retail brands under fake personas, inserting promotional links into every fifth post to appear organic. CIO Africa reported Anthropic found no evidence of direct government involvement, though the pro-government tone suggested possible alignment with the ruling coalition.
How the Campaign Spread Across Social Media
The operation, labeled GTG-54004, used hashtags like #PowerReliefKE and #PoweringTheNewKenya to amplify messaging. Posts praised the government's decision to withdraw Kenya Power's tariff-increase application in June 2026. Energy Cabinet Secretary Opiyo Wandayi said existing rates would remain after consultations to protect households and businesses from higher costs.
Beyond politics, the actor also ran a parallel operation promoting Kenyan retail brands. Posts were disguised as spontaneous customer reviews and organic recommendations, making coordinated promotion look like natural grassroots enthusiasm.
OpenAI's Tip Led to the Discovery
Zero Hedge reported that Anthropic's investigation began after OpenAI flagged repeat misuse of its platform. Anthropic then removed the account, developed detection measures for similar activity, and worked to prevent future campaigns using identical workflows.
The discovery is part of a broader pattern. Zero Hedge also noted that Anthropic has disrupted malicious campaigns from suspected state-sponsored groups in Russia and China, plus financially motivated criminals exploiting Claude for coordinated inauthentic behavior.
Why This Operation Ranked as Low-Level Threat
Anthropic classified the Kenyan operation as Category One on its Breakout Scale—the lowest tier. This means the campaign achieved limited impact beyond its original network and did not break through to mainstream visibility. The single actor's use of generic tactics and batch-posting kept the operation contained.
Still, the case highlights how AI chatbots can amplify political influence at scale. Mass-generating 50 coordinated posts mimicking organic opinion would have taken weeks by hand. Claude's speed enabled rapid, low-cost disinformation targeting Kenya's political landscape and consumer trust simultaneously.
Sentiment
Mostly NegativeNEGATIVENEUTRALPOSITIVE
Publishers
16
Articles
11
Reach
27
A
AnonymousAuthor
