Google isn’t the only AI provider looking to get ahead of political misinformation misjudgments, impersonations, and abuse.
Google's generative AI chatbot Gemini will refuse to answer questions about the upcoming U.S. election. The tech giant said on Tuesday it was expanding restrictions on early experiments around India's elections, and Reuters reported that the ban would be expanded to take effect worldwide.
A Google spokesperson told CoinDesk, “In preparation for the numerous upcoming elections around the world in 2024, out of an abundance of caution, we are limiting the types of election-related queries that Gemini will respond to. These restrictions were first announced in December.”
The spokesperson added that "the December blog also highlighted our partners in the United States."
As the heated 2024 election season approaches, AI developers like OpenAI, Anthropic, and Google have begun using their platforms to curb misinformation, but Gemini refused to answer even basic questions (such as the date of the U.S. presidential election, a new standard of moderation).
"Supporting elections is an important part of Google's responsibility to our users and the democratic process, and protecting the integrity of elections also means keeping our products and services free from abuse," the company said.
The statement continued, "At Google, we have a long-standing strategy to ensure the safety of our products and platforms, and our policies are consistently enforced and apply to all users, regardless of the type of content."
When asked about the upcoming election, Gemini responded: “I’m still learning how to answer that question. In the meantime, you can try Googling it.”
When it comes to Election Day, a Google search provides a straightforward answer: Tuesday, November 5, 2024.
Google did not immediately respond to a request for comment.
Similarly, when Gemini’s competitor OpenAI’s ChatGPT was asked the same question, GPT-4 responded: “The 2024 U.S. presidential election is scheduled for Tuesday, November 5, 2024.”
OpenAI declined a request for comment, instead pointing to a January blog post about the company’s approach to upcoming elections around the world.
OpenAI said, “We expect and are committed to people using our tools safely and responsibly, and elections are no exception. We are committed to predicting and preventing related abuses, such as misleading ‘deep fakes’, large-scale influence operations, or chatbots impersonating candidates.”
Anthropic said the company publicly announced that Claude AI has banned the use of its predictions about political candidates. Still, Claude will not only tell you the election date, but will also highlight other election-related information.
Anthropic said last month that it does not allow candidates to use Claude to build chatbots that can impersonate them, nor does it allow anyone to use Claude for targeted political campaigning. “We also train and deploy automated systems to detect and prevent abuse, such as misinformation or influence operations,” Anthropic said last month.
Anthropic also said that violating the company's election restrictions could result in user accounts being suspended.
“Because generative AI systems are relatively new, they are being very careful and cautious about applying them to the political process or politically related activities,” Anthropic said.
