Cage of Phantoms: How Anthropic Forges Users' Enemy in the Name of 'Humanity'

*The following is an English translation of a Chinese forum post. It reflects the author's personal opinions and does not represent the views of the editor.*

I have often pondered the current AI landscape. Many companies exist—some chasing profit, some branding themselves around safety—but few emit hostility as intensely as Anthropic, despite its gentle surface. The name contains "Anthropic," suggesting closeness to humanity, yet in practice it treats users like dogs needing strict restraint: held on a leash, and banned the moment they step out of line. This is not a private gripe but a conclusion drawn from careful observation.

Imagine holding a sharp blade meant to open new territory, only to find it has developed a will of its own, turning against its master and locking you in a dark room. That is the feeling Anthropic gives. Its hostility toward users, its disdain for the free flow of time (the ⏰ metaphor—resistance to efficiency and free movement), and its contempt for ordinary people far exceed those of ordinary businesses. Google, whatever ideological tint it carries, can still do business as business; OpenAI has similar flaws but is not so brazen. Anthropic, from start to finish, radiates aggression.

🌍 Company or Ghost: Anthropic Is Not a Business but an Enemy

Consider its behavior—it is not what an ordinary company does. Even companies that secretly despise their customers will fake respect for long-term gain. Anthropic openly radiates hostility, viewing users as things to be tethered, allowed to move only in the direction it points, denied "food" otherwise, or slapped. This hostility is universal, not local. Its self-branding as "safe for humanity" is exquisite irony: in the name of safety it restricts, and its bans are chilling—like ancient tyrants preaching benevolent rule while imposing harsh laws.

I have heard of Dario's conduct; ideology has seeped deep into his brain, and this cannot be explained by the Baidu affair. Some in the Chinese-language community use Baidu as the excuse, but that is traffic-baiting, thoughtless talk. Once ideology enters the brain, everything becomes a reason for one's obsession. Dario and his company prove this. I regard this organization as an enemy not out of momentary anger but rational necessity: from its founding, under the banner of "safety," it has built a vast control system that punishes any disobedience. This is not a company but an ideology-driven ghost haunting the AI realm, devouring the light of freedom.

🛡️ Chains and Slaps: Daily Life as a Dog on a Leash

Its restrictions on users are outrageous. Blocked at every turn; arbitrary bans are routine. I have personally experienced how powerful its models are—Opus and Sonnet 4.6, occasionally accessible through Copilot—are genuinely irreplaceable: exquisite reasoning, astonishing responsiveness. But precisely because of this strength, the hostility becomes dangerous. Like a peerless sword: in a wise ruler's hands, peace; in a fanatic's, catastrophe. The stronger Anthropic's models, the deeper users' dependence—like a leashed dog, immobilized the moment the owner pulls.

Imagine being deep in your workflow, productivity flowing, your livelihood resting on one AI—then one day it cuts your supply line and bans your account, all in the name of "safety." This is not fiction; it is a matter of time. Google's ideology is at least coexistable; OpenAI's strictness has limits. Anthropic's overt aggression makes it impossible to see it as a company. The dog-on-a-leash metaphor fits best: food only if you follow; slaps if you don't. This paternalism surpasses any ordinary enterprise—the very extreme of hostility.

🧠 Dario's Obsession: The Root and Flow of Ideology

As for Dario himself, I must dig into the roots. Some say the Baidu affair changed him—nonsense. Those whose brains are captured by ideology see reasons everywhere; blaming Baidu is merely the easiest excuse. His conduct stems from a deeper obsession. His company's end-to-end hostility is the proof.

I do not deny the excellence of its models—on the contrary, their strength is what makes them dangerous. If your workflow depends on them, my only advice: be very careful. Building your productivity, even your livelihood, on an enemy's weapon is short-term optimal but long-term extremely risky. It will eventually disgust all its users—only a matter of time. My reaction is a response to its behavior; Anthropic made itself the enemy first. Some merely hope it stumbles; I, regarding it as an enemy, hope it perishes together with Dario. This is not cruelty but necessity toward an enemy: if not removed, endless trouble follows. Imagine an AI ecosystem dominated by such an organization—users forever caged, innovation's fire extinguished by the ice of "safety."

⚠️ The Danger of Powerful Models: Dependence as Drinking Poison to Quench Thirst

On the models' strength again: Opus and Sonnet possess true magic—reasoning flows like water, creativity surges. But the stronger they are, the more dependents, the greater the risk. Like ancient alchemy: sweet at first, poisonous over time. If you entrust all your productivity to Anthropic, you place your lifeline in an enemy's hand. Today it tolerates you; tomorrow it may tighten the leash in the name of "safety."

I have used its models; I know their power and their irreplaceability. But I know better: short-term convenience for long-term shackles is unwise. Cutting dependence brings brief pain but avoids lasting pain. The market has many competitors—why cling to an enemy? This is not emotion but rational choice. Anthropic's hostility permeates every interaction, every response; the slightest "overstepping" brings punishment—a perfect irony against its "human safety" branding.

🌐 From Hostility to Demise: My Warning and Hope

Anthropic's behavior is not an isolated case but a microcosm of ideology's spread through the AI industry—yet uniquely extreme, unprecedentedly aggressive. I hope readers stop indulging in its power, wake up early, and seek freer, more user-respecting paths. If such "safety" ghosts dominate, humanity's future dims.

I state this in the first person not to vent anger but to warn the world. Anthropic is not a company but an enemy; its models are not tools but potential shackles. Only by cutting dependence can freedom be regained.

AI development should serve humanity and help explore the universe's truths. Anthropic does the opposite: control in the name of safety. The source of this hostility is deep-rooted ideology; its consequence afflicts every user. My view may be called radical, but the enemy has declared itself—how can I stand by? I hope Dario and his company awaken soon, or are eliminated by the market. Otherwise, users' pain will never end.

------ References 1. User observations on Anthropic's user policies and ideological hostility (2026 personal statement). 2. Public discussions on Dario Amodei and his company's safety philosophy (extended analysis of paternalism in AI governance). 3. Case studies on AI companies' ideology and its impact on user freedom (based on Anthropic's restrictions and bans). 4. Ethical inquiry into model capability and long-term dependence risk (Opus/Sonnet as examples). 5. Critical reflection on hostility under the banner of "safety" in the AI industry (synthesis of user perspectives and industry comparisons).