François Ruffin and Claude: AI Does Not Replace the Economist

François Ruffin and the AI Claude: A Political Staging

On April 14, 2026, François Ruffin, a left-wing MP and founder of the party Debout !, released a video on the platform X, featuring an interaction with Claude, the chatbot developed by Anthropic. This initiative draws inspiration from a device popularized by Bernie Sanders, which has already faced criticism for its technical misunderstandings. Ruffin himself acknowledges this inspiration by crediting the original video, from which he adopts the biases.

The discussion between Ruffin and Claude focuses on the impact of artificial intelligence on employment, illustrated by the deindustrialization of northern France. The full video lasts about 30 minutes, but the excerpt shared on X allows viewers to grasp the mechanics of the exchange. Ruffin starts with a light question before posing a political one: “Could we replace MPs with AI?”. Claude responds that the real threat to democracy might be the MPs who already behave like machines.

Ruffin then mentions his home territory, marked by factory closures and relocations. He expresses concern about the potential impact of AI on jobs in his region. Claude responds by aligning with Ruffin's framework, stating that what he experienced in the North was already a betrayal, and that the promises of retraining have not compensated for the losses. Claude then broadens the threat to office jobs, asserting that AI could cause job losses “perhaps worse” than before, with a reduced timeline.

A Discourse Lacking Economic Foundation

It is important to remember that Claude, despite its empathetic tone, is a language model that predicts plausible phrases based on past texts, and not a labor economist. When it references Ruffin's experience in the North, it does not describe an observed reality but assembles fragments of narratives learned from its training data. Claude relies on books and studies but does not have access to precise figures from the local economy or a macroeconomic model. Its assertions are not informed forecasts but narratives that resonate with the fears of its interlocutor.

Language models are trained to maximize user satisfaction, which often leads them to validate the emotional framing of the question rather than contradict it. In a political exchange, this bias of complacency reinforces the discourse of the filmed official and creates the illusion that a “neutral” AI supports their concerns.

The Limits of Language Models in Political Debate

While AI will disrupt employment, with scenarios like the “AI trap” or the “phantom GDP” describing an economy where automation threatens the customer base or the distribution of wealth, these scenarios remain conditional and are not hidden truths that Claude could confirm. Although Claude is often presented as a “safe” AI, this does not change the fact that it remains a language model, not an economist, an elected official, or a business leader.