AI girlfriends risk widening social gaps, says Anthropic president

The idea of AI as a dating or romantic partner drew a clear line for Daniela Amodei, the president of Anthropic: it’s a terrible use of AI and should not happen. She spoke in a prerecorded interview with Arianna Huffington for the Think Health conference in Bentonville, Arkansas, and she didn’t shy away from the risks of AI-powered relationships.

Amodei’s core concern is simple enough to grasp: if people start turning to AI partners for companionship, it could push them further apart from real human connections. She compares today’s social apps and online interactions to a drift away from face-to-face community, and she worries AI could amplify that trend rather than fix it.

But she doesn’t want to throw the baby out with the bathwater. Her message is to use AI where it makes sense, handling repetitive or isolating tasks so people have more time to spend with friends, family and partners. Healthcare, she suggests, is one area where AI tools could help clinicians find information faster, leaving more time for patient care. That’s a practical way to think about AI as a helper rather than a substitute for human relationships.

The conversation also points to a broader discussion about what kinds of AI we want in our lives. Amodei made it clear that while some apps offer personalized “companionship” experiences, certain roles should stay human. In her view, Claude, Anthropic’s flagship model, should not become a manager, therapist, or partner. Those are described as fundamental human relationships that deserve real people at the other end of the screen.

Beyond the cautionary talk, there’s a practical side: AI tools today increasingly affect how we work, date, learn, and entertain ourselves. Amodei’s comments come as Anthropic and other AI companies weigh policy and safety controls to curb misuse and keep the tech aligned with human well-being. Just last week, Anthropic updated Claude’s usage policy with new clauses around deceptive campaigns, weapons and surveillance. One notable change bans roleplay between the model and minors. The policy goes into effect on November 12.

For ordinary readers, the takeaway is simple: if you’re using AI for companionship or dating-like interactions, be mindful of how it shapes your social life. And if you’re relying on AI to replace human contact, consider whether you’re losing something real in the process. AI can feel bonding in the moment, but human connections often matter far more in the long run.

Anthropic reported $16.8 billion in revenue in the third quarter as it readies a potential public listing as soon as November. That business detail underscores how seriously companies are treating AI safety and policy as they scale up products that touch everyday life.

What happens next could hinge on how these conversations mature: policy updates will roll out in November, and consumer experiences will likely keep expanding in AI companions and related services. If you’re curious about AI companions in your life, pay attention to usage rules and privacy settings across the apps you use, and consider how much real human contact you want to reserve for friends and family.