Sciences Po Researchers Warn AI Will Shape 2027 Election, But With Limits
A note from Sciences Po researchers examines how AI could help both build and judge political programs, while stressing limits and democratic implications.
What happened
AI tools were studied for assisting program creation and citizen evaluation ahead of the 2027 election. Two experiments, IA Candidats 2027 and France Équilibre 2032, showed differences across models in priorities and budget ideas, with AI helping organize hypotheses but not guaranteeing economic relevance or political coherence.
Why it matters
The note highlights a new way to scrutinize programs, but also warns that data and estimates need verification and that political choices still rely on values.
The Sciences Po note argues that AI could both contribute to making political programs and help citizens judge them. It notes that AI might speed up research, compare scenarios, flag inconsistencies, and propose ideas, while stressing that outcomes depend on human checks and oversight.
Citizens could use AI to ask about pension, immigration, energy, or tax proposals and see what savings or beneficiaries might come from a policy. The report also says tools like wahl.chat already exist in Germany to compare party positions against their programs.
Despite the promise, the authors warn that AI alone cannot determine what is economically sound or politically coherent. Data must be checked and debated by experts, and choices depend on values beyond numbers.
What this does not tell us
Evidence is drawn from a single note and its described experiments; findings are exploratory and not validated as generalizable or causal.
FOR PEOPLE
Benefits reportedAI can help people compare and understand political programs without replacing judgment.
FOR AI AND ITS OPERATORS
Benefits reportedAI can organize information and show trade-offs, but cannot decide political truth.
These are two separate readings of what the sources describe. Reported claims and risks do not by themselves establish a real-world effect.
Original sources · 1
- Artificial Intelligence Tested by the 2027 Presidential Election ↗Sciences Po · 2026-11-06
Reporting discovered in Canada · France. Discovery market does not mean the event happened there.
What is your take?
Ask a question, add useful context or share a different perspective. Keep the conversation respectful and grounded.