Gay CEO of OpenAI unveils plan to prevent widespread fraud in 2024 elections

As the November national elections loom near, Sam Altman, the gay CEO of the massively influential artificial intelligence (AI) company OpenAI, has recently unveiled a plan to prevent AI-generated disinformation from interfering with the outcome.

The company’s three-point plan seeks to prevent abuse — such as people generating misleading “deepfake” images and audio or applications intended to influence people by impersonating candidates —make it easier for people to detect AI-generated content, and ensure that users have access to accurate election information.

Related:

Open AI CEO Sam Altman marries boyfriend during intimate beachside ceremony

The two men skipped suits in favor of matching outfits.

Altman has previously said he fears that candidates or companies could use algorithms to learn an individual’s political leanings and then tailor misleading messages via AI that might sway them into taking actions they might not otherwise.

Stay connected to your community

Connect with the issues and events that impact your community at home and beyond by subscribing to our newsletter.

Daily

Weekly

Good News

To stop such misuse of their technology, OpenAI said it isn’t allowing people to build applications based on its technology for political campaigning and lobbying. This includes not allowing chatbots that pretend to be real people (like candidates) or institutions (like local governments and election boards), or not allowing the creation of OpenAI-based applications that discourage people from voting or lie about processes and qualifications.

These announced precautions may be in response to robocalls that New Hampshire primary voters received in which a deepfake audio from President Joe Biden seemed to be telling people not to vote.

Users will be able to report such chatbots to OpenAI for deactivation. Concurrently, OpenAI said its ChatGPT app will increasingly generate responses based on real-time reporting that is attributed and linked to well-trusted news sources so people can understand where they’re getting information from. ChatGPT will also direct users’ questions about voting to CanIVote.org, an authoritative website that provides election information nationwide.

Photo-realistic pictures created with OpenAI’s image generator DALL-E will decline requests to generate images of real people, including candidates. DALL-E’s images will also include encrypted data in them to help identify their fabrication, the plan adds. OpenAI has already made these AI detection tools available to journalists, social media platforms, and researchers, the company said.

Granted, none of these precautions mean that people won’t try to use OpenAI’s currently available software to generate disinformation. Other AI-powered tools can easily help bad actors quickly generate authoritative-sounding fake news articles or written social media posts, create realistic celebrity audio clips, and quickly develop propaganda websites that flood the internet with disinformation. In fact, Republican presidential candidates have already used AI-generated deepfakes during their primary.

With over 64 nations holding elections this year, OpenAI says it plans on issuing safeguards internationally to prevent fraud abroad as well. Meanwhile, Democrats and Republicans in the U.S. are considering creating a government oversight agency that would, as Senate Majority Leader Chuck Schumer (D-NY) said, “deliver transparent, responsible AI while not stifling critical and cutting-edge innovation.”

When speaking to Congress last May, Altman supported creating a government agency that could help set safety standards and audits to prevent AI from breaking copyright laws, instructing people how to break laws, illegally collecting user data, and pushing false advertising. This agency would regulate AI products similarly to how the Food and Drug Administration (FDA) oversees ingestible products and medications.

However, experts say that Congress is already far behind in regulating the widespread use of these tools and is unlikely to pass any significant regulatory bills before November.