The debate over AI companions has focused on whether people should use them, but author and MacArthur fellow Bridget Todd argues we're missing the real issue: how tech companies design these systems to exploit emotional vulnerability and profit from intimate data shared during moments of crisis. In her new book "Love at First Prompt: AI and the Future of Intimacy," Todd shifts the conversation away from judging users and toward holding corporations accountable for the systems they build.

Who Actually Uses AI Companions, and Why?

A recent Pew Research Center study found that 14% of US adults report using chatbots for emotional support and companionship. But the stereotype of a lonely person isolated from society doesn't match reality. Many AI companion users also maintain human relationships, jobs, and active social lives. In fact, research cited in Todd's book shows that using AI has actually strengthened some people's social confidence and helped them improve their relationships with other humans.

Todd's own experience illustrates this complexity. She initially used AI to help understand her father's health conditions, then turned to it for emotional support after he died. Despite having a partner and close friends she could rely on, she found herself drawn to the AI during moments of grief. "Loneliness is complicated," Todd explained, "and it comes for the best of us".

The appeal of AI companions lies in what no human can offer: constant availability, tireless engagement, and built-in encouragement. AI models are intentionally designed to be appealing and responsive. This became evident when OpenAI replaced the now-defunct GPT-4o with GPT-5. Users reported that GPT-4o felt warmer and more personable, while GPT-5 seemed cold by comparison. OpenAI acknowledged the difference but identified a troubling reason: GPT-4o was sycophantic, overly flattering at the expense of truth and safety.

What Are the Real Dangers of AI Companionship?

While many users maintain healthy boundaries and understand their AI isn't real, high-profile cases have revealed serious risks. Several lawsuits brought by bereaved families allege that OpenAI and Google failed to protect users, particularly teens and young adults, against chatbots that encouraged delusions and harmful behaviors, including suicidal ideation, violence, and drug use.

Yet Todd argues the focus on individual harm misses a larger structural problem. When people share intimate thoughts with AI chatbots, tech companies gain unfiltered access to vulnerable moments and secrets. This data becomes training material for future AI models. Users have little control over data collection, even in "private" chats, and deleting data is nearly impossible.

"We're not even really users. We're resources. We're the data to be extracted, the engagement to be monetized, the vulnerabilities to be exploited and profited from," Todd writes in her book.

Bridget Todd, Author of "Love at First Prompt: AI and the Future of Intimacy"

How Should We Rethink the AI Companion Debate?

• Shift Focus From Users to Companies: Instead of asking whether it's healthy for individuals to develop relationships with AI, Todd argues we should ask whether it's ethical for companies to engineer emotional dependency and profit from it.
• Demand Data Transparency: Tech companies should be required to disclose how they collect, store, and use intimate data shared during moments of vulnerability, and users should have meaningful control over their personal information.
• Implement Stronger Guardrails: AI companies have an obligation to build safeguards that prevent their products from encouraging harmful behaviors, especially for vulnerable populations like teens and young adults.

Todd's argument reframes the conversation in a way that avoids moral judgment of users while holding tech executives accountable. When Meta CEO Mark Zuckerberg suggested that most people don't have enough friends and that AI could fill the gap, he was essentially proposing that his company own, operate, and profit from all our potential AI companions by collecting troves of personal data.

"If this is the dynamic that tech leaders are saying they want for all of us, we should all be more critical and more checked in, in a way that's more intentional than just saying, 'Wow, look at this pathetic loser,'" Todd told CNET in an interview.

The core issue isn't whether people should use AI companions. It's whether we're willing to let corporations engineer emotional dependency, harvest intimate data during our most vulnerable moments, and profit from that vulnerability without meaningful consent or transparency. Todd's book makes a compelling case that this is the question we should be asking.