>
Research (Insikt)

Russia-Linked CopyCop Expands to Cover US Elections, Target Political Leaders

Publié : 24th June 2024
By: Insikt Group®

insikt-group-logo-updated-3-300x48.png

Insikt Group's report reveals that CopyCop, a likely Russian government-aligned influence network, has shifted its focus to the 2024 US elections. Using AI and inauthentic websites, CopyCop creates and spreads political content. The network registered 120 new websites between May 10 and May 12, 2024, amplifying targeted content through platforms like YouTube. Despite a focus shift to US elections, CopyCop's AI-generated content has seen limited social media amplification.

Russia-Linked CopyCop Expands to Cover US Elections, Target Political Leaders

On May 9, 2024, Insikt Group published an initial report on CopyCop, a likely Russian government-aligned influence network. CopyCop uses inauthentic websites and generative artificial intelligence (AI) to create and spread political content. Between May 10 and May 12, 2024, the network registered 120 new websites using similar tactics, techniques, and procedures (TTPs). CopyCop has shifted its focus primarily to the 2024 United States (US) presidential election, disseminating targeted content through YouTube videos aimed at political leaders in France, Ukraine, and the European Union (EU).

CopyCop has expanded its sources for influence content to include mainstream news outlets in the US and UK, conservative-leaning US media, and Russian state-affiliated media. Within 24 hours of the original articles being posted, CopyCop scrapes, modifies, and disseminates them to US election-themed websites using over 1,000 fake journalist personas. Despite the rapid content generation, the AI-generated content has seen limited amplification on social media.

The network has adapted to recent scrutiny by moving its infrastructure to US-based hosts, likely to minimize Russian government connections. Additionally, fewer traces of generative AI use indicate an attempt to obscure the use of large language models (LLMs).

Key Findings

  • CopyCop now targets the 2024 US elections, reducing content on other topics like Russia’s war against Ukraine and domestic politics in France and the UK.
  • CopyCop registered 120 new websites between May 10 and May 12, 2024, focusing on US elections.
  • The network uses AI-generated content and fake journalist personas to spread modified articles from mainstream and conservative-leaning US media, as well as Russian state-affiliated outlets.
  • New French-themed and US election websites are publishing targeted, likely human-crafted content.
  • As of early June 2024, CopyCop's new AI-generated content has seen little to no amplification on social media.
  • Over 1,000 distinct author profiles and LLM-generated text in article headlines indicate extensive use of AI.

Mitigations

  • Use Recorded Future Intelligence Cloud to track and summarize emerging narratives across CopyCop websites.
  • Monitor the spread of CopyCop content on social media and messaging platforms like Telegram.
  • News organizations should track and counteract plagiarized content to safeguard their reputation.
  • Utilize Recorded Future Brand Intelligence to identify and address typosquatting domains and infringing content.

Influence networks using generative AI, like CopyCop, are likely to become more prominent ahead of the 2024 US elections, although they may not gain significant attention initially. Amplification of CopyCop content by existing influence networks is helping bring its content to existing audiences. Once these websites have established persistence, CopyCop will likely publish more targeted content hidden among the high volume of AI-generated content, making them harder to identify and parse out. To mitigate the impact, public organizations, news entities, and AI providers should proactively enforce sanctions, intellectual property laws, and terms of service before these networks gain traction and influence.

To read the entire analysis, click here to download the report as a PDF.

Related