In a new blog post, arguably the world’s most prominent AI company has shed light on its new and continued efforts to thwart election misinformation and disinformation via generative artificial intelligence.

One of the latest key areas the firm is exploring is image provenance, thereby helping the general public to know whether an image has been created by or doctored with AI, or if it’s legitimate.

OpenAI has been experimenting with a provenance classifier—a tool that can detect images generated by DALL·E, its AI text-to-image product. According to the company, internal testing of the tool “has shown promising early results,” even in instances where images have been further modified.

The firm also said that, early this year, it will implement DALL·E 3 with the Coalition for Content Provenance and Authenticity’s digital credentials, which encode details about where the image originated from via cryptography.

In terms of user limitations, subscribers of OpenAI’s tools aren’t allowed to create chatbots—like its new GPTs—that pretend to be real people or institutions like local government. Nor are they allowing users to build applications for political campaigning and lobbying, or applications that would deter people from electoral participation.

Meanwhile, ChatGPT itself will direct people asking US election-related questions, such as where they’re able to vote, to CanIVote.org—a non-partisan website created by state election officials that informs voters how and where they can register their votes.

“We’ll have more to share in the coming months,” the company wrote after sharing these details. “We look forward to continuing to work with and learn from partners to anticipate and prevent potential abuse of our tools in the lead up to this year’s global elections.”

Recommended reading

• OpenAI Launches GPT Store to Buy and Sell Custom Chatbots
• AI-driven Misinformation Tops WEF Short-term Threat List
• UNESCO Reveals Plan for Regulating Social Media Platforms

Despite OpenAI’s work on preventing election abuse, that’s not to say other AI tools—such as models that’ve been specifically created by bad actors for nefarious purposes—won’t be utilised in lieu.

The exploitation of AI for negative political purposes has emerged as a major, global concern, which the World Economic Forum highlighted last week. Through a poll of experts, the WEF found that the world’s largest short-term risk is AI-driven misinformation and disinformation.

In its following report, the NGO echoed fears that artificial intelligence in the hands of bad actors can be used to drive false narratives, incite conflict, and help foster climate denial.