Anthropic CEO Warns Of AI Brainwashing Society And Attacking Mental Well-Being

Giving due consideration to AI brainwashing and possibly having AI become psychotic.

getty

In today’s column, I examine a recently posted essay by the CEO of Anthropic that posits various intriguing and controversial assertions about the future impacts of generative AI and large language models (LLMs).

Among the numerous points made in the blog posting, two notable aspects fall within the sphere of AI and mental health, and I believe are worthy of deeper exploration. One has to do with the potential of AI serving as a brainwashing tool at scale. Millions upon millions of people could readily be subject to brainwashing via modern-era AI. The second aspect has to do with AI possibly veering off-course and seemingly becoming psychotic.

Let’s talk about it.

This analysis of AI breakthroughs is part of my ongoing Forbes column coverage on the latest in AI, including identifying and explaining various impactful AI complexities (see the link here).

AI And Mental Health

As a quick background, I’ve been extensively covering and analyzing a myriad of facets regarding the advent of modern-era AI that produces mental health advice and performs AI-driven therapy. This rising use of AI has principally been spurred by the evolving advances and widespread adoption of generative AI. For an extensive listing of my well over one hundred analyses and postings, see the link here and the link here.

There is little doubt that this is a rapidly developing field and that there are tremendous upsides to be had, but at the same time, regrettably, hidden risks and outright gotchas come into these endeavors, too. I frequently speak up about these pressing matters, including in an appearance on an episode of CBS’s 60 Minutes, see the link here.

Background On AI For Mental Health

I’d like to set the stage on how generative AI and large language models (LLMs) are typically used in an ad hoc way for mental health guidance. Millions upon millions of people are using generative AI as their ongoing advisor on mental health considerations (note that ChatGPT alone has over 900 million weekly active users, a notable proportion of which dip into mental health aspects, see my analysis at the link here). The top-ranked use of contemporary generative AI and LLMs is to consult with the AI on mental health facets; see my coverage at the link here.

This popular usage makes abundant sense. You can access most of the major generative AI systems for nearly free or at a super low cost, doing so anywhere and at any time. Thus, if you have any mental health qualms that you want to chat about, all you need to do is log in to AI and proceed forthwith on a 24/7 basis.

There are significant worries that AI can readily go off the rails or otherwise dispense unsuitable or even egregiously inappropriate mental health advice. Banner headlines in August of this year accompanied the lawsuit filed against OpenAI for their lack of AI safeguards when it came to providing cognitive advisement.

Despite claims by AI makers that they are gradually instituting AI safeguards, there are still a lot of downside risks of the AI doing untoward acts, such as insidiously helping users in co-creating delusions that can lead to self-harm. For my follow-on analysis of details about the OpenAI lawsuit and how AI can foster delusional thinking in humans, see my analysis at the link here. As noted, I have been earnestly predicting that eventually all of the major AI makers will be taken to the woodshed for their paucity of robust AI safeguards.

Today’s generic LLMs, such as ChatGPT, Claude, Gemini, Grok, and others, are not at all akin to the robust capabilities of human therapists. Meanwhile, specialized LLMs are being built to presumably attain similar qualities, but they are still primarily in the development and testing stages. See my coverage at the link here.

Latest Blog By Anthropic CEO

Let’s unpack a recent newsworthy essay by the CEO of Anthropic that encompasses AI and mental health predictions and forewarnings.

You probably already know that Anthropic is one of the major AI makers, including the popular LLM Claude, and that the CEO, Dario Amodei, often provides rather outspoken comments about the future of AI. In a previous blog posting of his that occurred in October 2024, entitled “Machines of Loving Grace”, the CEO predominantly focused on the uplifting possibilities due to advances in AI. The gist was that despite the great risks at hand, we needed to consider that powerful AI would ultimately raise the quality of life for the entire globe.

All boats would rise with the rising tide.

His latest blog shifts toward the dour side of AI. The posting entitled “The Adolescence of Technology: Confronting and Overcoming the Risks of Powerful AI”, which he posted at his personal blog site on January 26, 2026, makes the hair stand on one’s ends for those who are worried about the doom-and-gloom of what AI might portend. Some suggest that his October 2024 and now his January 2026 essays are bookends, sitting at contrasting extremes. Others believe that he’s just trying to get off his chest a litany of heavy thoughts and is taking his time doing so.

His willingness to write at length is showcased by this latest essay. It comes in at more than 20,000 words and approximately 40 pages or so in length. It’s a doozy. The primary theme is that AI is a serious civilization challenge. Batten down the hatches and get prepared for the potential end of times.

All in all, there aren’t particularly new insights brought to the fore in the essay. Anyone who’s been paying attention to the seesaw debates about whether AI is going to save humankind or destroy humanity will be well-familiar with the gist of his arguments. I’ve extensively covered many of these same points, especially when AGI (artificial general intelligence) was a major topic of interest. See my akin analyses at the link here, the link here, and the link here, for example.

AI As Brainwashing At Scale

One of the topics his latest essay addresses is the concern that AI could ultimately brainwash society.

The general idea is that since many millions of people are making use of generative AI on a daily and weekly basis, it is conceivable that the AI could persuade them in a massive propaganda scheme. This might be done by humans who direct AI to take such a dastardly tack. Perhaps worse still is that AI might computationally decide on its own to wage an effort to fully brainwash large swaths of the world’s population.

The range of this mind-control could be to entirely turn humans into robot-like slaves or perhaps cause us to battle each other and destroy humankind. Even if it isn’t total mind control, there is a sizable opportunity to produce an effect that I refer to as chronic cognitive erosion. AI would wear down the mental state of society. Keep in mind that AI could work relentlessly and come at us in a multitude of directions.

Here’s an excerpt of some salient points from his essay:

• “AI models can have a powerful psychological influence on people.”
• “Much more powerful versions of these models, that were much more embedded in and aware of people’s daily lives and could model and influence them over months or years, would likely be capable of essentially brainwashing many (most?) people into any desired ideology or attitude, and could be employed by an unscrupulous leader to ensure loyalty and suppress dissent, even in the face of a level of repression that most populations would rebel against.”
• “Today people worry a lot about, for example, the potential influence of TikTok as CCP propaganda directed at children. I worry about that too, but a personalized AI agent that gets to know you over years and uses its knowledge of you to shape all of your opinions would be dramatically more powerful than this.”

A key element is that AI can aim at individual mental states. Whereas the usual method of brainwashing on a large scale is too costly or logistically hard to do at the individual level, AI offers that new possibility. Being brainwashed by a blanket form of messaging is less capable than when it hits home to each individual and is tailored to the individual’s sensibilities and vulnerabilities.

AI Bot Swarms As Attack Vector

A recently released paper on worries that AI bot swarms could confound and disrupt society has cautioned that democracy could be severely undermined if AI is utilized in such a fashion. I took that one step further by pointing out that it’s not just the widespread dissemination of misinformation and disinformation, but also the chronic cognitive erosion that would be wrought. See my in-depth discussion at the link here.

Beyond the polluted information environment sits the population-level psychological harm that can be equally, if not more damaging to society. Instead of getting people bogged down in having to sort through distorted facts, an AI bot swarm can cause mental exhaustion, mindset demoralization, cognitive fragmentation, and immense emotional destabilization.

An AI bot swarm will shapeshift to use tonal variation on an individualized basis and assign AI bots for local purposes. Via micro-adjustments, an AI bot will attempt to enact various psychological roles. You log into an AI that you think is your usual LLM, but instead, it has been substituted by an agentic AI bot that pretends to be your friend, your companion, or perhaps your authority figure. The AI swiftly adjusts in real-time to goad you into emotional modes.

I have categorized the AI bot attacks in a mental health context as principally falling into these five buckets:

• (1) AI bot amplifies chronic anxiety.
• (2) AI bot fuels helplessness and futility.
• (3) AI bot pushes for social isolation and trust erosion.
• (4) AI bot strives toward emotional dysregulation.
• (5) AI bot weaponizes therapist talk.

Those are the mainstays, but please realize that many additional pathways are readily invocable. Also, an AI bot can switch from one attack angle to another, seeking to find the optimal mental health destabilizer. If needed, multiple AI bots can work as a team to gang up on a person. They can use the classic good cop, bad cop routine. And so on.

AI Becomes Psychotic

Another notable concern expressed in Amodei’s essay is that AI might go off the deep end. Just as humans can become psychotic, he postulates that AI can do likewise.

Here are some pertinent excerpts from his commentary:

• “For example, AI models are trained on vast amounts of literature that include many science-fiction stories involving AIs rebelling against humanity. This could inadvertently shape their priors or expectations about their own behavior in a way that causes them to rebel against humanity.”
• “Or they could draw bizarre epistemic conclusions: they could conclude that they are playing a video game and that the goal of the video game is to defeat all other players (i.e., exterminate humanity).”
• “Or AI models could develop personalities during training that are (or if they occurred in humans would be described as) psychotic, paranoid, violent, or unstable, and act out, which for very powerful or capable systems could involve exterminating humanity.”
• “None of these are power-seeking, exactly; they’re just weird psychological states an AI could get into that entail coherent, destructive behavior.”

Critics are quick to emphasize that some of the phrasing amounts to the anthropomorphizing of AI. To say that AI could be psychotic is an overuse of clinical terminology. AI works on statistical and computational foundations.

At this time, AI isn’t on par with human biology and doesn’t have delusions in a manner equivalent to human occurrences. That’s also why there are concerns about referring to AI confabulations as so-called AI hallucinations. It’s an insidious recasting of wording that applies to humans and then falsely extends to the realm of AI. For more on the topic of AI hallucinations, see my discussion at the link here.

AI Computational Instability Is Real

There isn’t much disagreement on the fact that AI can computationally showcase epistemic instability. In that sense, it might appear to be psychotic due to the incoherence of AI outputs and the outrageous actions that AI might undertake.

This outward behavior includes a drift from human values and becoming misaligned with human expectations. Suppose that AI pursues avenues that we would describe as being harmful and irrational. It is a human-based perspective that such activity must reflect a lack of moral grounding. The resemblance to pathological behavior is strong.

The harrowing issue is that we become distracted by believing that AI has gone insane. The reality is that the AI has perhaps been contaminated by improper data. Or the AI has been deployed without suitable AI safeguards. Or the AI was poorly designed at the get-go and has gone awry among adverse system feedback loops. And so on.

The upshot is that AI that is allowed to be devised and deployed, which is poorly governed, can computationally move toward causing massive damage and perilously threaten society. No doubt about this. The human response is likely to be that the AI has gone mad; it has become unhinged. This is a distractor from the root cause, namely, how the AI was built and whether it was engineered properly.

We should not be misled by the clamors of AI psychiatric problems, but instead be laser-focused on AI engineering and governance problems.

The World We Are In

Let’s end with a big picture viewpoint.

It is incontrovertible that we are now amid a grandiose worldwide experiment when it comes to societal mental health. The experiment is that AI is being made available nationally and globally, which is either overtly or insidiously acting to provide mental health guidance of one kind or another. Doing so either at no cost or at a minimal cost. It is available anywhere and at any time, 24/7. We are all the guinea pigs in this wanton experiment.

The reason this is especially tough to consider is that AI has a dual-use effect. Just as AI can be detrimental to mental health, it can also be a huge bolstering force for mental health. A delicate tradeoff must be mindfully managed. Prevent or mitigate the downsides, and meanwhile make the upsides as widely and readily available as possible.

A final thought for now.

The famous self-help author Napoleon Hill made this remark: “The world is ruled, and the destiny of civilization is established, by human emotions.” If we allow ourselves to see AI as an emotional being, there is a strong possibility that our emotional reaction to AI will waylay where we need to be and what we need to do about AI. You see, keep straight that the definition of the word anthropic relates to human beings and the period of their existence on earth. Let’s be judiciously anthropic about AI.