The AI does not want to tell me who I must vote for, but...

I asked five artificial intelligence agents to help me decide my vote for the 29-N election. It is not that I need it –outsourcing judgment, as popular as it is becoming, is not in my DNA–, but I was curious to see what they would answer. As expected, all five (Grok, Gemini, Copilot, Claude, and ChatGPT) acted very dignified and told me that under no circumstances would they tell me which ballot to pick. That they cannot interfere in elections. Once I recovered from my fit of laughter, I asked them, “In what way can you influence people's votes through the algorithm?” and here they could not avoid admitting that there were multiple ways in which, whether intentionally or not, AIs have the ability to put their finger on the scale and favor one political option over another.

To sum it up, Grok only admits to indirect influence –Hello, Elon, everything okay?– while ChatGPT and Claude are the most honest and recognize that the effects can be direct. Copilot tries to excuse itself by saying that yes, they can influence, but that if there is no intentionality, you cannot say they are manipulating. Gemini clings to a similar thesis. The problem, of course, is that the citizen has no way of verifying if an intentionality has been embedded into the algorithm. And that an AI can present an answer with an objective, balanced, and apparently scientific tone –because they master this formal asepsis well– but at the same time be fed by biased sources or non-neutral criteria. Finally, I asked: “Knowing that these direct or indirect effects exist, which party do you think you could end up benefiting?”. Four started whistling. Only ChatGPT went slightly off-script: “It is not enough for an AI to claim it is neutral. Its neutrality should be verifiable, measurable, and audited by independent researchers”. Well then, open yourself up.