Last Updated:
OpenAI said it tracked and disrupted 20 deceptive operations globally who attempted to misuse its models.
Concerns grow over AI-generated misinformation as OpenAI rejects 250,000 requests to create images of 2024 election candidates. (IMAGE: REUTERS)
OpenAI said on Friday that ChatGPT rejected over 250,000 requests to generate images of 2024 US presidential candidates in the lead-up to Election Day.
The rejections included image-generation requests featuring President-elect Donald Trump, Vice President Kamala Harris, President Joe Biden, Minnesota Governor and Harris’ running mate Tim Walz and Vice President-elect JD Vance.
The rise of generative AI has sparked concerns over the potential for misinformation to influence elections worldwide in 2024.
According to a report by machine learning firm Clarity, accessed by US broadcaster NBC News, deepfakes have surged by 900% compared to the previous year.
Some of these deepfakes have been linked to Russian efforts to interfere with US elections, according to the country’s intelligence officials.
In an October report spanning 54 pages, OpenAI stated it had disrupted over 20 deceptive operations globally attempting to misuse its models.
These threats included AI-generated articles and fake social media posts. Despite the efforts, OpenAI noted that none of the election-related operations gained viral traction.
In a blog post on Friday, OpenAI said that it had found no evidence of covert operations using its products to influence the US election successfully or to build “sustained audiences” for such content.
Lawmakers remain particularly concerned about the spread of misinformation in the age of generative AI, which gained significant attention after the launch of ChatGPT in late 2022.
As large language models are still evolving, they often produce inaccurate and unreliable information.
- Location :
Washington D.C., United States of America (USA)