To Reason or Not To Reason: LLMs and the Streamlining of Influence Operations

Competing papers from Apple and Anthropic debate whether LLMs are capable of reasoning, while malign actors continue using them to shift their influence efforts into hyperdrive.

and

Jun 19, 2025

∙ Paid

Fight Russian disinformation and click the ❤️ heart above to help Substack’s recommendation algorithm show this post to more people. 🙏🏻☀️

In 2015, disrupting the information space required substantial resources: rented offices for troll farms, dozens of operatives coordinating fake accounts, trained language specialists, and IT infrastructure. This sort of thing was a coordinated, fairly resource-intensive operation. Today, in 2025, the same scale of manipulation can be achieved by a single person using a ChatGPT prompt and an AI image generator. The barrier to entry into the world of sowing distrust and division in democracies has collapsed, and with it, the dynamics of influence have fundamentally changed.

The American Sunlight Project

To Reason or Not To Reason: LLMs and the Streamlining of Influence Operations

Competing papers from Apple and Anthropic debate whether LLMs are capable of reasoning, while malign actors continue using them to shift their influence efforts into hyperdrive.

This post is for paid subscribers