Google AI Mode withholds direct answers on recent US election winners
Original title: Google AI gets cagey when asked who won the 2020 elections
The same questions get straight answers abroad, on a second try, and from Gemini, ChatGPT, and Claude.
Ars Technica reports that Google AI Mode often answers recent US presidential-winner questions with links instead of a direct result, while Gemini and other major chatbots answer plainly. VPN checks from Taiwan, Russia, France, and Canada, and repeated questions, produced straightforward answers citing Google Civics data. AI Mode attributed the first response to strict safety filters that send users to official sources such as the Federal Election Commission. A Google spokesperson said the Gemini app answers more fully, called the Search response unhelpful, and said Google is working to address it.
Unconfirmed: AI Mode's account of its safety filters is the model's own explanation, which the report treats as unreliable.
- Foreign VPN checks in Taiwan, Russia, France, and Canada returned direct answers.
- A second ask produced winners for three elections previously answered only with links.
- YouTube in 2023 reversed a ban on false 2020-election videos.
- Google's September post said it would highlight official midterm resources.
- Vendors
- Organizations
- OpenAIAnthropicFederal Election Commission
- Countries
- United States
From the original article
2 min read at arstechnica.com
But the AI responses get more forthright when asked outside the US.
By Kyle Orland · arstechnica.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.