We explore why Google is restricting election-related answers from its AI model, Gemini, and what this means for our daily lives.
Imagine this: You’re having a busy morning and ask your smartphone AI assistant, “Who were the candidates in the last presidential election I saw on the news yesterday, and what policies did they propose?” Instead of an answer, the AI says, “I am still learning how to answer that question. Why don’t you try Google Search instead?”
Many users of Google’s AI chatbot, ‘Gemini,’ have recently experienced this situation. The artificial intelligence, which usually answers everything, is keeping its mouth shut specifically when it comes to ‘elections’ or ‘political results.’ Why has Google placed these restrictions on an AI known for being so smart?
Why Does This Matter?
In the past, when we had questions, we would flip through encyclopedias or seek out experts, but now we live in an era where we ask AI. The information provided by AI directly influences our thoughts and judgments. If an AI gives the wrong answer or presents a one-sided opinion on a sensitive topic like an election, the ripple effect could be massive.
Google applies strict guidelines to election-related queries to prevent such risks in advance. This can be viewed as a ‘safety fence’ to prevent AI from generating ‘fake news’ and causing confusion in the election process, which is the foundation of democracy.
Easy to Understand: AI’s ‘Basic Education’ Process
In simple terms, Gemini not answering election questions is not because it ‘doesn’t know,’ but because it has been ‘decided not to.’
It is easier to understand if you compare it to a ‘filter’ on a camera app. Just as there are filters that emphasize or remove certain tones when taking a photo, Google has installed an ‘election-related answer filter’ in Gemini’s brain.
AI develops by learning from massive amounts of data, and this data is mixed with countless political views and unverified claims. If an AI generates answers based on this data probabilistically, there is a risk that it could unintentionally convey biased information as if it were the truth. That is why Google chose to guide the AI to look up verified official sources (e.g., Google search results, Federal Election Commission, etc.) rather than having the AI provide probabilistic answers on sensitive topics like elections [Source: Google restricts global election query answers on Gemini, Source: Google, Google restricts AI chatbot Gemini from commenting on global election queries].
Current Situation: Different Appearances for Different AI Services
Interestingly, even with the same Google technology, the response method varies slightly depending on the type of service. In some tests, while Google’s AI-powered search feature directed users directly to the National Archives or pages of election management agencies, Gemini, an independent chatbot, sometimes answered directly that Joe Biden was the winner [Source: Google AI mode won’t say who won 2020 election… what about Gemini?].
While Google’s internal policy is firm, discrepancies appear in the actual implementation across different services. This clearly shows that AI technology is not yet perfectly controlled and is undergoing adjustments and improvements in real-time [Source: Google’s Gemini AI Chatbot Restricted from Global Election Queries].
How Much Should We Trust It?
The tension between AI and election information will continue. As technology advances, AI will provide more sophisticated answers, but the possibility of generating misinformation that could harm the fairness of elections will not disappear.
We are gradually becoming accustomed to the convenience of asking AI everything. However, when important decisions are required, we should not blindly trust AI’s answers. As Google recommends, we should refer to AI’s guidance but develop the habit of cross-verifying facts through direct searches. After all, AI is just a smart assistant, and the final decision-maker must always be ‘us.’
MindTickleBytes AI Reporter’s Perspective
Silencing AI may be a wise measure in the short term to prevent the spread of misinformation. However, a question remains: is an AI that doesn’t provide answers truly an ‘intelligent assistant’? Finding the balance between technical safety and the usefulness of information provided is the biggest task that AI companies must overcome in the future.
References
- To increase the AI's processing speed
- To prevent the spread of misinformation and fake news
- To reduce the revenue of Google's search engine
- Generates and provides a direct answer
- Ignores the question and talks about another topic
- Guides the user to use Google Search or similar resources instead of answering directly
- All AI services provided the same answer
- The AI search feature guided users to relevant institutions, while the chatbot sometimes answered that Joe Biden won
- Both services refused to answer the question