DiResta: AI makes influence ops cheaper, not more persuasive

Renée DiResta's new Lawfare essay argues that generative and agentic AI have made influence operations faster and cheaper to run without making their audiences easier to reach. "Generative AI scales content. Agentic AI scales operations," DiResta, an Associate Research Professor at Georgetown's McCourt School of Public Policy, writes. The scale she describes on the production side, in her reading, has not translated into visible persuasion.

Her specifics come largely from abuse reports the model providers have been publishing. Iranian propaganda institutions, she writes, had Claude "produce doctrine manuals and ministerial planning documents that, in Anthropic's assessment, would otherwise have required a fully staffed program office." A French advertising firm ran a pipeline feeding roughly 70 fake news sites that published nearly 9,000 articles in about 20 languages and "drew almost no real engagement." MEK operators had an AI agent study about 8,400 of a real activist's Telegram posts, then hold live conversations masquerading as him with his contacts inside Iran.

The historical benchmark DiResta sets against those campaigns is the Russian "troll factory" whose "Project Lakhta" monthly budget "exceeded $1.25 million" by September 2016, according to a Justice Department indictment. That operation worked, she argues, less because of its volume than because American newsrooms covered the material it surfaced.

Her bottom line: "AI can make an influence apparatus cheaper, faster, and more persistent, but it cannot manufacture trusted distributors." Two of the analysts we follow in AI Weekly's Who's Who had already posted the piece on the day it went up.