OpenAI Sets AI Guardrails for US, Global Elections Integrity

The U.S. election season has officially begun with the first primary elections concluding in Iowa this week. Given the growing concern about the role of AI in spreading misinformation and interfering with the democratic process both in the U.S. and abroad, OpenAI says it has taken steps to prevent any misuse of its services.

The ChatGPT maker has published a blog post outlining the measures it plans on taking to ensure that its platform is not misused. Among them include DALL-E users being unable to generate images of political candidates. Users also cannot use the new GPT Builder tool to build applications for political campaigning and lobbying.

“Protecting the integrity of elections requires collaboration from every corner of the democratic process, and we want to make sure our technology is not used in a way that could undermine this process,” the Microsoft-backed startup said.

New image detection tools

OpenAI is implementing a way to tell whether an image was generated by DALL-E 3.

OpenAI plans to use digital credentials from the Coalition for Content Provenance and Authenticity, which encode details about the content’s provenance using cryptography.

OpenAI is also experimenting with a provenance classifier for detecting DALL-E images and plans to make it available to journalists, platforms and researchers for feedback. However, no timeframe was given as to when this would occur.

Related:OpenAI in Talks with CNN, Fox to License Content

New ChatGPT integrations

OpenAI’s blog post states that ChatGPT is getting integrations from news sites. Users will get access to real-time news reporting including links to full stories.

The integrations stem from OpenAI’s deals with publishers. It has currently secured agreements with Politico, Business Insider and the Associated Press, and in talks with others including CNN, Fox and Time.

OpenAI said that transparency around the origin of the news information “can help voters better assess information and decide for themselves what they can trust.”

OpenAI also revealed it is working with the National Association of Secretaries of State (NASS) to provide users with information on where they can vote. When prompted with election-related queries, ChatGPT will direct users to CanIVote.org, the website for U.S. voting information.

“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,” OpenAI said.