The creator of the innovative artificial intelligence platform, OpenAI, has announced its intention to deploy measures to hinder the spread of false information in anticipation of numerous pivotal elections scheduled this year across nations housing nearly half the global populace.
The phenomenal rise of the sophisticated AI text-generating platform, ChatGPT, has propelled a worldwide craze in the field of AI, yet it has also raised concerns over the potential for such technologies to inundate the web with misleading content, potentially influencing the electorate.
With a series of critical polls on the horizon in nations such as the United States, India, and the United Kingdom, OpenAI declared on Monday its stance against the utilization of its creations—including ChatGPT and the visual creation tool DALL-E 3—in the sphere of political canvassing. The firm is determined to ensure its innovations do not detrimentally impact the integrity of the democratic process, as stated in a recent company blog entry. OpenAI acknowledged that it is presently evaluating the persuasive capabilities of its tools and maintains a policy of prohibiting the development of applications specifically for political advocacy and voter solicitation.
The World Economic Forum recently highlighted the issue of AI-abetted false narratives as one of the most significant and immediate global threats, with the potential to destabilize governments just coming into power in leading nations. Concerns regarding electoral misinformation aren’t new, but the advent of highly efficient AI-driven tools for generating text and images has exacerbated these worries, as noted by specialists. This is particularly true when it’s challenging for the populace to discern between authentic and altered content.
OpenAI disclosed on Monday its ongoing work on mechanisms that will authenticate text produced by ChatGPT, while also empowering individuals to detect whether an image originates from DALL-E 3. In the early part of the year, the company plans to adopt the digital credentialing standards of the Coalition for Content Provenance and Authenticity (C2PA), employing an encryption-based method to embed traceable information about digital content’s origins. This coalition, whose participants include tech heavyweights like Microsoft, Sony, Adobe, and camera magnates Nikon and Canon from Japan, focuses on enhancing means to track digital media.
To instate ‘guardrails,’ OpenAI said its ChatGPT, when prompted with procedural voting queries in the U.S., such as polling locations, will redirect users to official sources. The firm indicated that insights from these measures would shape its protocols for additional territories and regions.
The construction of images depicting actual individuals, including political contenders, is thwarted by inbuilt safeguards in DALL-E 3. This move by OpenAI comes on the heels of strategies laid out the previous year by U.S. tech behemoths Google and Facebook’s parent company Meta to stem election meddling, particularly via the application of AI.
In past fact-checking efforts, AFP had debunked deepfakes—manipulated footage—claiming to show U.S. President Joe Biden proclaiming a military draft as well as a faux endorsement of Florida Governor Ron DeSantis by former Secretary of State Hillary Clinton. Ahead of Taiwan’s presidential race, AFP Fact Check also encountered social media traffic showing altered audio and video of politicians, though the quality was generally poor and it was uncertain if AI was employed. Even if AI is not always the tool used, experts are pointing to misinformation as a driver of a growing crisis of confidence in public institutions.