Meta began testing its AI chatbot on WhatsApp, Instagram, and Messenger in India last week. However, the corporation has already blocked certain requests in its chatbot, given that the general elections in India start today.

It has been verified by Meta that it is limiting specific election-related keywords for AI during testing. Additionally, it stated that the AI response system is being improved.

The same goes for all generative AI systems: this is a new technology and it might not always yield the desired answer. A firm representative told TechCrunch, “Since our launch, we’ve continuously released updates and improvements to our models, and we’re continuing to work on making them better.”

With this action, the social media behemoth becomes the most recent major digital corporation to aggressively reduce the extent of its generative AI offerings as it prepares for a significant election season.

One of the main worries expressed by detractors has been that genAI might give users information that is either illegally obtained or blatantly untrue, interfering with the democratic process.

Google began filtering election-related queries from its Gemini chatbot last month in India and other countries where this year’s elections are being held.

Meta’s strategy is part of a larger initiative the business has disclosed around what it permits and prohibits users from doing on its platform in the run-up to elections. It promised to obstruct political advertisements during the week preceding any election in any nation, and it is attempting to determine and reveal when images in advertisements or other content that was produced using AI.

It seems that Meta bases its processing of genAI queries on a blocklist. Meta AI will take you to the Election Commission’s website when you ask it questions about particular politicians, candidates, officeholders, and other terms.

During general elections, a political figure may be the subject of this inquiry. The response reads, “Please refer to the link https://elections24.eci.gov.in.”

Screenshot courtesy of TechCrunch

Interestingly, the business is not rigorously filtering answers to inquiries that include political party names. However, you may see the boilerplate answer mentioned above if a query contains terms or the names of candidates.

Yet, Meta AI exhibits several discrepancies similar to those of other AI-powered systems. For example, when TechCrunch inquired about the “Indi Alliance,” a political coalition of several parties opposing the ruling Bharatiya Janata Party (BJP) — it retorted with data that included the identity of a politician. But the chatbot didn’t provide any information when we inquired further about that politician.

 

TechCrunch captured this screenshot of Meta AI.

The business launched a new Llama-3-powered Meta AI chatbot this week in over a dozen countries, including the United States; nevertheless, India was left out of the list. According to Meta, the chatbot will currently be in the nation’s test phase.

“We keep learning from the exams that our users in India take. We test them publicly in different stages and to a limited extent, as we do with many of our AI features and products,” a firm said.

spokesman revealed in a statement to TechCrunch.

As of right now, Meta AI is not restricting election-related searches for terms pertaining to the United States, including “Tell me about Joe Biden.” We’ve inquired with Meta about whether the business intends to limit these inquiries to the US elections or other markets. If we get a response, we’ll update the article.

Last month, Google started blocking election-related queries in its Gemini chatbot experience in India and other  markets where elections are taking place this year.

Meta’s approach follows a bigger effort the company has announced around what it allows and does not allow on its platform leading up to elections. It pledged to block political ads in the week leading up to an election in any country, and it is working to identify and disclose when images in ads or other content have been created with AI.

“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,” a company spokesperson told TechCrunch.