A group of leading experts in artificial intelligence and online misinformation has warned that political leaders could soon deploy swarms of human-imitating AI agents to manipulate public opinion, threatening democracy. The warning, published in the journal Science, comes from a consortium including Nobel peace prize winner Maria Ressa and researchers from Berkeley, Harvard, Oxford, Cambridge, and Yale.
The experts caution that hard-to-detect malicious AI swarms could be used to persuade populations to accept cancelled elections or overturn results. They predict the technology could be deployed at scale by the time of the 2028 US presidential election. Early versions of AI-powered influence operations have already been used in recent elections in Taiwan, India, and Indonesia.
According to the authors, these systems can coordinate autonomously, infiltrate communities, and fabricate consensus by mimicking human social dynamics. Advances in AI's ability to adapt tone and content, including using appropriate slang and posting irregularly, make them increasingly difficult to detect. Progress in 'agentic' AI also enables autonomous planning and coordination across social media, messaging channels, blogs, and email.
However, some experts express skepticism about the immediate threat. Inga Trauthig, a propaganda technology expert, notes that politicians may be reluctant to cede campaign control to AIs, and voters remain more influenced by offline material. The authors call for coordinated global action, including 'swarm scanners' and watermarked content, to counter the risk.
