To counter AI-driven misinformation, OpenAI’s ChatGPT rejected 250,000 deepfake picture requests of candidates within the month earlier than the 2024 election.
What Occurred: Final week, OpenAI stated ChatGPT turned down 1 / 4 of 1,000,000 requests to create deepfake photographs of candidates utilizing DALL-E, the corporate’s AI artwork generator.
ChatGPT was additionally programmed to reply logistical queries about voting by directing customers to CanIVote.org, a U.S. voting data web site run by the Nationwide Affiliation of Secretaries of State.
See Additionally: Google Gemini Rolls Out ‘Utilities’ Extension For Android: Right here’s What All You Can Do With It
ChatGPT offered roughly a million responses directing customers to the voting web site in the month main as much as Nov. 5.
On Election Day, ChatGPT was set to reply questions on election outcomes by referring customers to respected information organizations just like the Related Press.
“Round 2 million ChatGPT responses included this message on Election Day and the day following,” the platform said in a weblog publish.
Subscribe to the Benzinga Tech Tendencies e-newsletter to get all the most recent tech developments delivered to your inbox.
Why It Issues: The announcement adopted rising fears that AI might disrupt the marketing campaign by producing deepfakes and conspiracy theories for on-line unfold.
In January, New Hampshire voters obtained robocalls with a deepfake voice of President Joe Biden, urging them to not vote within the state’s main.
Earlier this 12 months, the Middle for Countering Digital Hate raised considerations concerning the misuse of AI picture creation instruments from OpenAI and Microsoft Corp. for election-related disinformation.
Later in August, OpenAI uncovered and dismantled a covert Iranian affect operation leveraging ChatGPT to control public opinion in the course of the 2024 elections.
By October, OpenAI had thwarted over 20 international operations and misleading networks that sought to misuse its fashions for election interference, in keeping with a 54-page report revealed by the corporate.
Try extra of Benzinga’s Client Tech protection by following this hyperlink.
Learn Subsequent:
Disclaimer: This content material was partially produced with the assistance of AI instruments and was reviewed and revealed by Benzinga editors.
Picture courtesy: Shutterstock
Market Information and Knowledge delivered to you by Benzinga APIs
© 2024 Benzinga.com. Benzinga doesn’t present funding recommendation. All rights reserved.
