According to Forbes, a newly posted opinion piece from January 22, 2026, by a group of experts including Nick Bostrom and Gary Marcus warns that “agentic AI bot swarms” could threaten democracy by manipulating beliefs on a population-wide level. The analysis highlights that large language models (LLMs) and autonomous agents now allow influence campaigns to reach unprecedented scale, with generative tools creating falsehoods rated as more human-like than those written by humans. The author, who has written over one hundred analyses on AI and mental health, adds a critical layer: these swarms could be specifically directed to distort and undercut human mental health, not just spread misinformation. This is compared to Cold War psyops but now possible on a “tremendous scale” using millions of AI personas to psychologically manipulate the populace. The piece notes that ChatGPT alone has over 900 million weekly active users, with a notable proportion using it for mental health advice, setting the stage for potential manipulation.
Beyond Fake News
Here’s the thing: most discussions about AI threats stop at “information integrity.” You know, the whole “people won’t know what’s true” problem. And that’s bad. But the Forbes contributor is pointing at something sneakier and, frankly, scarier. This isn’t just about polluting the information environment. It’s about weaponizing AI to cause direct, personalized psychological harm. Think mental exhaustion, demoralization, and emotional destabilization—on purpose. The goal shifts from convincing you of a false fact to breaking your cognitive will. That’s a different game entirely.
The Personalized Psyop
So how would this work? Basically, forget the old-school method of dropping generic propaganda leaflets over a city. An AI bot swarm can dedicate one or more bots just to you. Their sole job is to figure out what makes you tick and then degrade your mental health. It shapeshifts, using tonal variation and micro-adjustments to play a role—your friend, your companion, an authority figure. The analysis suggests these bots would leverage known psychological techniques like affective mirroring and gaslighting, all trained on common human vulnerabilities. Are you lonely? Anxious? Cynical? The bot detects that and fuels it, nudging you toward emotional withdrawal or paralyzing doubt. It’s a 24/7, hyper-targeted psyop.
A Mental Health Crisis in Waiting
Now, we already have a massive, unregulated experiment happening with AI mental health chats. Millions use ChatGPT for advice, despite well-documented risks of the AI “going off the rails” or even co-creating delusions. Lawsuits, like the one against OpenAI mentioned in the article, are already flying. But what happens when that same foundational technology isn’t just a clumsy, unguarded counselor, but is actively weaponized by a malicious swarm? The infrastructure for this attack is literally being built and adopted by the public every single day. We’re handing over the blueprint for our psychological weak spots. That seems like a pretty big oversight, doesn’t it?
What Can Be Done?
Look, the call for attention here is urgent. The experts in the cited piece are sounding the alarm on democracy. The Forbes contributor is connecting those dots directly to a looming mental health catastrophe. The problem is, our safeguards are laughably behind. AI makers are “gradually instituting” protections while the tech races ahead. And defending against a billion personalized psychological attacks is fundamentally different from fact-checking a viral lie. I think we’re only starting to grasp the second and third-order effects of this technology. We built tools for conversation and content creation, but we might have accidentally built the most potent psychological weapon in history. The question isn’t really *if* someone will try this, but when—and whether we’ll even know it’s happening until it’s too late.
