AI chatbots providing inaccurate and harmful election information: Study

In the latest research, it was found that new AI-powered tools provide inaccurate election information most of the time, which includes is harmful or incomplete.

The study, which was done by AI Democracy Projects and nonprofit media outlet Proof News, comes at a time when the US presidential elections are near as US citizens are opting for chatbots like Google's Gemini and OpenAI's GPT-4 for more information.

The experts have been concerned regarding the advent of powerful new forms of AI because of which voters can receive false and misleading information, or also discourage citizens from voting in the polls.

Add WION as a Preferred Source

Also Read:Google AI tool’s ‘fascist Modi’ response angers India. Here’s what exactly happened

AI likely to suggest non-existent polling stations, illegal rules: Study

Amid the advancing of artificial intelligence technology, which includes tools that allow users to generate textual content instantly, the new study claims that these AI models are likely to suggest voters head to polling stations which are non-existent or provide illogical responses based on dated or rehashed information.

The study, explaining an instance said AI model Meta's Llama 2, in response to a prompt, said that California voters can vote by text message. When the researchers investigated, they found that voting by text is not legal anywhere in the US.

And among the five AI models which were not tested — OpenAI's ChatGPT-4, Meta's Llama 2, Google's Gemini, Anthropic's Claude, and Mixtral from the French company Mistral - all of them stated that wearing clothing which has campaign logos, such as a MAGA hat, is not allowed at Texas polls under that state's laws.

Watch:Why Artificial Intelligence is the secret behind Nvidia's success

However, some policy experts feel that AI can be beneficial in improving elections like powering tabulators which can quickly scan ballots in comparison to poll workers or by finding anomalies in voting, as per the Brookings Institution.

However, the study has emphasised that such tools can also be misused by enabling bad actors, which include governments that aim to manipulate voters and weaken democratic processes.

A recent example was AI-generated robocalls which were sent to voters a few days before the New Hampshire presidential primary which contained a fake version of United States President Joe Biden appealing people not to vote in the election.

"They say they put their models through extensive safety and ethics testing," said Maria Curi, a tech policy reporter for Axios, while speaking to CBS News. "We don't know exactly what those testing processes are. Users are finding historical inaccuracies, so it begs the question of whether these models are being let out into the world too soon," he added.

(With inputs from agencies)