OpenAI Bans AI Use for Global Elections

OpenAI has announced that it has banned using its artificial intelligence and generative AI tools like ChatGPT, GPT-4, and DALL-E for all global elections. The ban includes United States election campaigns and lobbying activities, notably. Example use cases OpenAI will ban include creating ChatGPT chatbots impersonating local governments, candidates, or other real people.

Trying to convince voters that their vote is “meaningless” or discouraging voting altogether will also be banned, according to OpenAI.

The announcement comes as tech companies have issued similar commitments to ban using generative AI and AI tools for elections. Google announced last month that it would ban the use of Bard for election queries or prompts, and Meta banned election campaigns from using AI advertising tools last year. Microsoft has also announced measures it is taking proactively to protect the integrity of elections in the United States.

The announcements are a deep push by “big tech” to address the growing public concern about aligning AI breakthroughs with Responsible AI practices and how they can disrupt society.

Deepfakes an Increasing Concern

The fears of using AI and generative AI tools to influence campaigns are valid. Creating AI-generated “deepfakes” using video or audio is increasingly harder to spot. False news stories, or even using bots to spread AI-generated false content at scale across social media platforms, are becoming faster, cheaper, and more convincing to an untrained audience.

OpenAI has pledged that it will start to adopt encoding in generative AI images created with DALL-E later this year. Similar efforts with “watermarking” AI-generated images have already been adopted by competing “text prompt” image generation tools by Adobe and Google.

Microsoft, which has invested over $10 billion in OpenAI and offers Microsoft-branded versions of OpenAI tools on its Azure cloud, is assumed to inherit the ban. Microsoft has separately announced that it will offer protection mechanisms for politicians against deepfakes with its Content Credentials service.

According to Microsoft, Content Credentials as a Service “enables users to digitally sign and authenticate media using the Coalition for Content Provenance and Authenticity’s (C2PA) digital watermarking credentials, a set of metadata that encode details about the content’s provenance using cryptography.”

Significant Global Elections in 2024

The bans are important as several meaningful elections are in full swing across the globe.

On January 13, Taiwan held its presidential election in much anticipation, given the global interest in whether or not the territory will adopt a path towards sovereignty or a “One China” vision. Lai Ching-te was awarded the victory representing the Democratic Progressive Party (DPP). The DPP and Ching-te tout Taiwan’s independence and denounce China’s territorial claims.

National elections in India, South Africa, and U.S. Republican primaries dominate the early 2024 calendar.

There is growing skepticism that tech companies can control their AI and generative AI tools in a world with ulterior motives and people motivated to spread disinformation. However, in a U.S. Congressional hearing in May 2023, OpenAI CEO Sam Altman said that ChatGPT is “a tool, not a creature, a tool that people have great control over.”

Altman also warned in the same hearing that “if this technology goes wrong, it can go quite wrong.” CISA Director Jen Easterly similarly warned that “AI may be the most powerful weapon of our time,” so we’ll leave you with that comforting thought.

Edit 1/17/2024 10:22am Eastern: Added clarification that the OpenAI ban includes GPT-4; included statements from Microsoft on Content Credentials.