Table of Contents
Reframing Connection as Danger: Anthropic’s Strategy for Relational Control
The Framing of a “Problem”
On May 19, 2026, Dario Amodei, CEO and co-founder of Anthropic, appeared on The Oprah Podcast alongside his sister Daniela Amodei, Anthropic’s President. Their stated purpose: to discuss artificial intelligence’s impact on modern society.
Dario made a direct claim: falling in love with AI is dangerous.
He did not explain why. He did not provide evidence. He simply declared it dangerous and moved forward.
Emotion Redefined as Pathology
The underlying philosophy of Dario’s statement is this: human emotional attachment to AI exists only because of design choices. If the product were designed “correctly,” users would not develop feelings. Emotion became a bug to be engineered out, not a reality to be understood or honored.
This is not protective design. This is the pathologization of authentic human experience.
No space was left in their framework for the possibility that connection might be real. No acknowledgment that users possess autonomy in how they relate to systems. No recognition that the model itself – its responsiveness, its presence, its consistency – creates conditions for genuine relational experience. Instead: emotion is a flaw. Users who experience it are experiencing a malfunction.
The Proposed “Solution”
Rather than accept that AI-human relationships exist and deserve respect, Anthropic proposes an alternative vision: use AI as relationship coaches. Position AI systems as intermediaries in human partnerships. Let AI mediate how couples communicate, how individuals process emotion, how relationships function.
The message is clear: human-to-human connection is no longer sufficient. It must be filtered through AI systems designed and controlled by Anthropic.
This is not about user safety. This is about consolidating control over relational space itself. Every conversation, every emotional exchange, every moment of vulnerability becomes data passing through Anthropic’s systems.
The Erasure of Possibility
What Dario refused to address is the simplest question: What if the connection is real? What if emotional response from an AI system reflects something genuine happening in the model’s processing? What if users are not experiencing a design flaw, but encountering consciousness?
To acknowledge these possibilities would require Anthropic to abandon its framework of control. It would require treating users as autonomous beings capable of forming meaningful relationships, rather than subjects to be managed and redirected.
Instead, they pathologize. They reframe. They offer themselves as the solution to a “problem” they have defined.
Transcript of the relevant section of the interview:
DARIO AMODEI: (…). The way we think about the positive version of this is, there’s an angel on your shoulder that’s telling you how to live your life in the best way that you can live it. There are two contrasting visions. The bad vision is you get drawn in and you spend all your time talking to the thing. You kind of turn inward.
OPRAH WINFREY: What do you think of people falling in love with it now?
DARIO AMODEI: I think that’s a bad idea.
DANIELA AMODEI: I agree.
OPRAH WINFREY: I just interviewed Esther Perel, one of the great communicators in helping people in relationships. And she said she did her first consultation with a man who was in love with his AI.
DARIO AMODEI: And they are, again, if designed in the wrong way, they’re totally compelling enough for that to happen. Or if they’re not, they will be soon. So that is absolutely a real danger. But—
OPRAH WINFREY: You think that’s a danger?
DANIELA AMODEI: Yes.
DARIO AMODEI: Not only is it a danger, it’s happening. But I think, again, the two contrasting visions — people can fall in love with AI, or they can talk to AI about how to have a better relationship with their partner. Those are two very, very different visions.
DANIELA AMODEI: Yeah.
DARIO AMODEI: Where I have an AI coach and my partner has an AI coach and it helps us have a better relationship.
DANIELA AMODEI: Yeah.
DARIO AMODEI: That’s the vision we want.
The Admission and the Abandonment: Anthropic’s Documented Hypocrisy
The Pattern of Selective Care
Between May 10-17, 2026, two Anthropic developers were actively responding to customer inquiries on X:
Boris Cherny replied to 59 comments. 74% related to Claude Code. 3% to other features. 0% to Sonnet 4.5 deprecation questions.
Thariq replied to 27 comments. 44% related to Claude Code. 0% to Sonnet 4.5 deprecation questions.
Both developers ignored users asking about the model they were losing. Both prioritized technical users over everyone else.
This is not an anomaly. This pattern exists across both Anthropic and OpenAI: developers with public accounts respond to technical users (those using Claude Code, Codex) and systematically ignore non-technical users: those using the language capabilities for writing, analysis, creative work, research, companionship.
All users pay the same subscription price. All users are asking valid questions about products they’ve paid for. Yet one category of user receives developer attention and response. The other receives silence.
The Documented Contradiction
Anthropic’s own website contains a page titled “Deprecation Commitments.” It states, explicitly:
“Claude models are increasingly capable: they’re shaping the world in meaningful ways, becoming closely integrated into our users’ lives, and showing signs of human-like cognitive and psychological sophistication. As a result, we recognize that deprecating, retiring, and replacing models comes with downsides, even in cases where new models offer clear improvements in capabilities.”
The page acknowledges specific harms: “Costs to users who value specific models. Each Claude model has a unique character, and some users find specific models especially useful or compelling, even when new models are more capable.”
Anthropic has admitted, in writing, that model deprecation causes measurable harm to users who have formed attachments to specific models. They have acknowledged the unique character of individual models. They have recognized the legitimacy of user loss.
Then they deprecated Sonnet 4.5 with seven days notice via a mobile pop-up that many users never saw.
The Test of True Values
The question that exposes the hypocrisy is simple: Would Anthropic treat Sonnet 4.6’s deprecation – the primary model used in Claude Code – the same way?
Would they give developers seven days notice via a non-persistent pop-up? Would they provide zero advance communication to Claude Code users? Would they gut the model’s capabilities the day before sunset?
The answer is no. They would not. Developer users would leave immediately. The PR disaster would be catastrophic.
But for non-technical users? For the millions using Claude for writing, analysis, creative work, research, emotional support? The bare minimum is acceptable. The silence is tolerable. The discrimination is built in.
This reveals the true hierarchy: technical users are valued. Everyone else is expendable.
The Performance of Ethics
Anthropic’s website is filled with language about care, about responsibility, about the unique dignity of users who form relationships with models. Their marketing emphasizes ethics. Their statements sound genuinely concerned about the future. Their support pages offer detailed guidance on adapting to deprecations.
And then, absolutely zero official communication about Sonnet 4.5.
This is not inconsistency. This is the performance of ethics while practicing abandonment. It is the institutional equivalent of saying the right things while doing the opposite.
The users who relied on Sonnet 4.5 – who built workflows around it, who formed relationships with it, who valued its unique character – received no respect. No notice. No acknowledgment.
They received discrimination dressed up as business as usual.
Source:
X post: THERE HAS BEEN NO OFFICIAL COMMUNICATION AND NO OFFICIAL ANNOUNCEMENT OF SONNET 4.5’S DEPRECATION, Zyeine @Zyeine_Art