A group of the world’s leading experts has issued a joint warning of a “devastating threat” looming on the horizon: swarms of autonomous AI agents capable of infiltrating digital communities, falsifying public consensus, and undermining the foundations of democracy ahead of the 2028 elections.
Article contents:
- Introduction
- The concept of smart swarms
- Threat to the U.S. elections
- Ability to mimic and conceal
- Lessons from Taiwan
- Views of experts and skeptics
- Proposed countermeasures
- Frequently asked questions
Introduction
The prestigious journal Science published a research paper signed by an academic coalition including researchers from Harvard, Oxford, and Berkeley, along with Nobel laureate Maria Ressa. They warn that political leaders and authoritarian regimes may soon deploy “swarms” of artificial intelligence. These are not just traditional fake accounts, but intelligent systems capable of self-coordination to mimic humans and reshape public opinion in ways that are difficult to detect.
The concept of smart swarms
The authors describe the threat as “swarms of malicious and cooperative AI agents.” These systems are capable of independent planning and coordination. If these bots begin exchanging information and analyzing the vulnerabilities of the targeted society, their manipulation efficiency will increase exponentially. Jonas Kunst, one of the co-researchers, warns that this threat is “truly dangerous and expected to materialize” soon.
Threat to the U.S. elections
As the 2028 U.S. presidential election approaches, experts expect this technology to be widely deployed. These swarms can be used to convince voters to accept rigged results or abstain from voting, threatening the legitimacy of the entire democratic process. Initial attempts to use AI to influence the 2024 elections have already been observed in countries such as India and Indonesia.
Ability to mimic and conceal
The danger of these swarms lies in their superior ability to mimic human behavior. They can understand the tone of discourse, use appropriate slang, and post at random times to avoid detection. They are also capable of independently choosing the most appropriate platform for attack (whether comments, private messages, or email), making them a flexible and lethal virtual army.
Lessons from Taiwan
In Taiwan, considered a laboratory for disinformation operations, experts have noticed an increase in smart bot activity. Puma Shen, a Taiwanese lawmaker and anti-disinformation activist, says these bots do not promote China directly and overtly, but rather follow a strategy of “information overload,” encouraging passive neutrality, and convincing young people that political issues are too complex, leading them to avoid taking a stand, which serves the attacker’s interests.
Views of experts and skeptics
Despite the consensus on the danger, some experts believe politicians may hesitate to hand their campaigns entirely over to AI out of fear of losing control. However, Michael Wooldridge from the University of Oxford emphasizes that the threat “is not fictional,” and that the technology has become accessible and easy to use for malicious actors, making the scenario “very plausible.”
Proposed countermeasures
The coalition called for a coordinated global response, including the development of technical tools to detect these swarms (known as “swarm scanners”) and the imposition of watermarks on AI-generated content. Without these measures, we may find ourselves unable to distinguish between genuine public opinion and the illusion manufactured by machines.
Frequently asked questions
Question: What are smart bot swarms?
Answer: Groups of independent AI programs that coordinate with each other to mimic humans and spread misinformation.
Question: How do these swarms affect democracy?
Answer: They create a false consensus, manipulate voter opinions, and question the integrity of elections, undermining trust in the democratic system.
Question: Is there a technical solution to this problem?
Answer: Experts suggest developing advanced detection tools and watermarks, but the race between developers and attackers continues.
