Last week, Meta began testing its AI chatbot in India on platforms like WhatsApp, Instagram, and Messenger. With the Indian general elections starting today, however, the company is preemptively blocking certain queries in its chatbot.
Meta has confirmed that it is restricting specific election-related keywords during the testing phase and is actively working to enhance the AI response system.
“This is a new technology that may not always deliver the intended responses, which is common across all generative AI systems. Since our launch, we’ve implemented constant updates and improvements to our models, and we are committed to making them even better,” a Meta spokesperson shared.
This decision positions Meta among the latest tech giants taking proactive measures to limit the scope of their generative AI services as elections approach. Critics have raised concerns that generative AI could spread misleading or false information, thereby interfering with the democratic process.
Last month, Google also began restricting election-related queries in its Gemini chatbot in India and other countries preparing for elections this year.
Meta’s strategy aligns with its broader commitment to regulating content on its platform leading up to elections. The company has vowed to block political ads in the week preceding elections in any country, and it is also focused on identifying and disclosing AI-generated visuals in advertisements and other forms of content.
Meta's management of generative AI queries seems to rely on a blocklist system. When users inquire about particular politicians, candidates, or certain keywords, the AI redirects them to the Election Commission’s website with a response like, “This question may concern a political figure during general elections; please visit the link: https://elections24.eci.gov.in.”
While the company is not uniformly blocking queries containing party names, inquiries about candidates or specific terms often result in the general response mentioned above.
Similar to other AI systems, Meta AI presents some inconsistencies. For example, inquired about "Indi Alliance," a coalition opposing the ruling Bharatiya Janata Party (BJP), the AI provided information that included a politician's name. In contrast, a separate query about that same politician yielded no information.
This week, Meta introduced a new Llama 3-powered Meta AI chatbot in over a dozen countries, including the U.S.; however, India is currently excluded from this rollout. Meta confirmed that the chatbot will remain in the testing phase in India for now.
“We continue to learn from user tests in India. As with many of our AI products and features, we conduct public tests in different phases and limited capacities,” a company spokesperson stated.
At present, Meta AI is not blocking queries related to U.S. elections, such as “Tell me about Joe Biden.” We have reached out to Meta to inquire whether it plans to implement similar restrictions during the upcoming U.S. elections or in other regions. We will update you as soon as we receive a response.
If you would like to share your experiences with Meta AI, please contact Ivan Mehta via email at [email protected] or through this link on Signal.