20261012 Chatbots beurteilen Wahlkandidaten

12.10.2026 Chatbots beurteilen Wahlkandidaten

Chatbots manipulate Users

Josh Kalla, a political scientist at Yale University, spent two years studying how chatbots can influence people’s political opinions. And this applies even to people who are fully aware that they are talking to a computer. Currently, many U.S. citizens are asking their trusted chatbot about the upcoming elections — the U.S. midterms.

In September, the Institute for Strategic Dialogue (ISD) published a report revealing serious quality issues with chatbots — regardless of which company had made the service available to the public. Nearly one-third of the responses to election-related questions from six AI chatbots tested were incomplete, inaccurate, or outdated, according to Telepolis.de.

The criticism covers the following issues:

Chatbots continue to produce hallucinations. They simply invent “facts” without citing sources.
The AI systems are completely opaque. No one knows what data they were trained on or how they were programmed.
No one knows what criteria they use to evaluate the data available to them.
Anthropic and Google did not respond to inquiries; OpenAI cited security precautions when asked about voting behavior and claims that its chatbot behaves “neutrally.”

It seems the era of voter autonomy will soon be over. This makes it clear that the heads of Big Tech could play a key role in the upcoming elections.