OpenAI Combats Election Misinformation Amid Growing Concerns

ChatGPT maker OpenAI is racing to deter the use of its artificial intelligence models in spreading election misinformation as lawmakers warn of increased threats from online influence campaigns throughout the 2024 voting cycle.

See Also: Live Webinar | Integrating Splunk and Panther for Real-Time Alerting and Custom Dashboarding

The AI powerhouse announced in a Monday blog post a series of steps it will take to improve user access to authoritative voting information and prevent the use of ChatGPT to discourage people from voting.

“We expect and aim for people to use our tools safely and responsibly, and elections are no different,” the post read. The company detailed how it has restricted users from building applications for political campaigning and lobbying, or creating chatbots that can impersonate real-life candidates, political institutions or local governments.

The announcement comes amid increasing fears that political deepfakes and AI-generated election misinformation could significantly disrupt democracy. A majority of election leaders in the United States are most concerned about threats to election security stemming from disinformation campaigns, which can be significantly more effective with the help of generative AI tools, according to a January survey of more than 130 state and local government officials conducted by the Center for Digital Government.

More than 45 countries are set to cast votes this year, including the United States, the European Union and likely the United Kingdom. The World Economic Forum calculates elections held this year will determine leadership in countries that produce more than half of the world’s gross domestic product (see: Generative AI Concerns Grow in Record Election Year).

On Wednesday, U.S. Sen. Mark Warner, D-Va, chairman of the Senate select committee on intelligence, wrote a letter to Jen Easterly, director of the Cybersecurity and Infrastructure Security Agency, urging her to recommit to addressing critical election threats, including both foreign and online influence operations.

The letter stated that election threats have continued to grow in recent years, “with a wider array of foreign actors” and “a larger number of social media platforms suitable for influence activity.”

OpenAI said it was experimenting with several new tools that can “empower voters to assess an image with trust and confidence in how it was made.” The company plans to implement a new approach this year created by the Coalition for Content Provenance and Authenticity to embed digital credentials in content created by its image generation tool, DALL-E 3.

Users can also report potential violations to OpenAI, which said in the blog post that it will be releasing a new tool for detecting images generated by DALL-E to an initial subset of testers, including journalists, researchers and other platforms for feedback.

The company said it was working with the nonpartisan U.S. National Association of Secretaries of State and planning to direct users to CanIVote.org when asked specific questions about voting and the 2024 election.

“As we prepare for elections in 2024 across the world’s largest democracies, our approach is to continue our platform safety work by elevating accurate voting information, enforcing measured policies, and improving transparency,” the blog read.

Five American states have already enacted laws regulating the use of deepfakes in elections, while similar bills are making their way through legislative houses in at least 18 other states, according to tracking information from the non-profit organization Public Citizen.