Next year's elections in the UK and US could be marred by a wave of AI-driven disinformation, experts have cautioned. Generative AI tools like ChatGPT and Midjourney enable the creation of convincing text, images, and deepfake videos at scale, potentially amplifying propaganda through swarms of AI-powered bots.
Sam Altman, CEO of OpenAI, told a US congressional hearing that AI models could manipulate users. 'The general ability of these models to manipulate and persuade, to provide one-on-one interactive disinformation is a significant area of concern,' he said. Altman advocated for regulation, public education, and corporate responsibility to address the risks.
Prime Minister Rishi Sunak announced on Thursday that the UK would lead efforts to limit AI dangers. Prof Michael Wooldridge of the Alan Turing Institute described AI-powered disinformation as his primary concern, noting that generative AI can produce tailored falsehoods targeting specific voter groups, such as Conservative supporters in the home counties or Labour voters in metropolitan areas.
Voice cloning technology has already been misused, as seen in a doctored video of President Joe Biden that used ElevenLabs' tool to simulate his voice. The company has since tightened safeguards. Cybersecurity firm Recorded Future reported that rogue actors are selling voice cloning services online, including the ability to impersonate corporate executives and public figures.
While some, like Altman, draw parallels to early Photoshop, others warn that advancing AI makes it increasingly difficult to trust online content. With elections approaching, the threat of interactive disinformation at scale looms large.
