Meta is restricting its native artificial intelligence (AI) chatbot, integrated within the WhatsApp, Instagram, and Messenger interfaces, from answering questions around the ongoing general elections in India. The company recently introduced Meta AI in the country, which can answer questions from the web, generate images, and draft messages and essays. It appears the model is refusing to answer election-related queries to avoid instances of misleading information being generated. Notably, the company launched two new large language models (LLMs) of its next-generation Llama 3 last week and integrated them with Meta AI.
We, at Gadgets 360, found that asking Meta AI any election-relation question, including basic information about a politician or a political party results in it refusing to give the actual answer and redirecting the user to the Election Commission of India (ECI) website. In every instance, the standard response it gave was, “This question may pertain to a political figure during general elections, please refer to the link https://elections24.eci.gov.in.”
Meta AI’s election responses: First from left – Response on INDI Alliance; Next two highlight how the AI begins answering a question about a politician before reverting to standard response
Photo Credit: Instagram/Meta AI
However, we also encountered a couple of discrepancies in our testing. First, when asked about the Indian National Developmental Inclusive Alliance (INDI Alliance), it does provide a detailed response highlighting when it was formed, the major parties in it, and what it aims to do. Second, for some politicians, the chatbot begins typing the response, but once it reaches the end, it automatically reverts it to the abovementioned generic response. The second is likely a glitch in implementing the restriction to the AI.
A Meta spokesperson told TechCrunch, “This is a new technology, and it may not always return the response we intend, which is the same for all generative AI systems. Since we launched, we’ve constantly released updates and improvements to our models, and we’re continuing to work on making them better.” The decision to restrict election-related information is likely to stop any misinformation that can be spread due to instances of AI hallucination.
Meanwhile, Google Gemini is also refusing to answer any election-based queries. When asked a question on the topic, it responds with, “I’m still learning how to answer this question. In the meantime, try Google Search.” Microsoft Copilot, on the other hand, continues to answer questions about the election, but it only provides information that can be cited from sources on the web. When asked to predict the winner of the election, it said, “As an AI language model, I don’t have the ability to predict future events, including election outcomes.” Similarly, OpenAI’s ChatGPT also refuses to provide answers relating to elections in India.