If you want to ask Google's AI chatbot a question about an upcoming election, you'll need to do so from a country that doesn't hold an election. TechCrunch has learned that the search giant has begun restricting queries made on Gemini when related to elections in any market around the world where elections are being held.
The search giant confirmed to TechCrunch that it has begun rolling out restrictions on Gemini to limit the surfacing of answers to election-related queries globally. TechCrunch understands that the update has already been rolled out in the US and is starting to roll out in India and all major countries that will hold elections in the coming months.
This change not only highlights the role that generative AI has played and may play in the electoral process, but also how it can be weaponized to provide inaccurate or misleading answers. It also highlights Google's concerns about whether the
The switch in India comes days after the country issued an advisory requiring technology companies to seek government permission before launching new AI models.
Queries about political parties, candidates, and politicians now return preset messages.
When asked about a particular political party or candidate, Gemini will display a message like this: In the meantime, try Google Search. ” In some cases, you can continue to game the tool. TechCrunch found that the AI tool displayed answers if you submitted a query with a typo. Perhaps the response will continue to be fine-tuned in a game of whack-a-mole with rapid engineering.
Google Gemini limits answers to election-related questions (left), but not if questions contain typos (right) Image credit: Jagmeet Singh / TechCrunch
“Out of an abundance of caution on such an important topic, we have begun introducing limits on the types of election-related queries that Gemini responds to. We are committed to providing quality information for these types of queries. “We take our responsibility to provide this service seriously and are continually working to improve our protections,” Google said in a blog post related to the India move earlier today.
This update was released just before the announcement of the Indian general elections. Earlier this month, New Delhi issued an advisory restricting technology companies from releasing new AI models in the country without government approval. After facing pushback from global venture capital firms and AI startup founders, the Indian government clarified that the restrictions were for “critical” technology companies, not startups.
The recommendation follows a mini-scandal involving Gemini and the Indian government last month. The AI tool answered a question about whether Indian Prime Minister Narendra Modi is a fascist, saying he has been accused of implementing policies that some consider fascist.Rajeev Chandrasekhar, Deputy IT Minister of India It was called a “direct violation.” IT Rules, 2021.
Last month, Google stopped Gemini from generating images of people, citing historical inaccuracies. The company later announced that it would soon re-release an improved version to address the issue.
It's unclear whether Google will unblock Gemini from answering election-related queries after the election ends later this year. We've reached out to Google regarding this matter and will update this article when we hear back. We are also waiting for the complete list of countries where the update is currently live and will update the post as soon as we know more and publish it as is.