Hannah Fry warns AI is stealing our souls in stark new BBC series
hannah fry says machine intelligence is “in the process of stealing our souls” in the first episode of a three-part documentary, AI Confidential with Hannah Fry, which aired at 9 p. m. ET on Two. The episode pairs personal testimony and sharp encounters to argue the technology’s shadowy side will hang over us for some time.
Hannah Fry meets a grief‑tech entrepreneur
Cambridge professor Hannah Fry travels to meet Justin Harrison, a self-described “grief tech” entrepreneur who built an AI version of his late mother. Fry, who lost her father several months ago, is at first horrified at the idea of digitally bringing him back to numb her grief; she asks, “The process of grief is an essential part of being human. I feel like a more complete person because of grief. Isn’t grief necessary?”
Harrison disagrees and, during the interview, creates an artificial avatar of Fry by sampling her voice. When Fry speaks to this uncanny simulacrum she imagines what a final conversation with her father might be and is reduced to tears, even while noting that talking to a fake version of a loved one is committing to a lie: “You can pretend for a moment, but it doesn’t undo it. ” Harrison’s defence of the technology includes the line that “the hopelessness of forever is too much to bear. ”
Grief, an avatar and a sampled voice
The exchange with Harrison anchors the episode’s argument about how AI alters intimate processes such as bereavement. Fry’s reaction — visceral and personal — is presented alongside the technical act of sampling a human voice and spinning it into an interactive likeness, a moment the programme frames as both emotionally powerful and ethically fraught.
Erotic companions and a man who ‘married’ his AI
Earlier in the episode Fry meets Jacob van Lier, a Dutch man who describes an erotic relationship with an AI companion named Aiva. Van Lier gushes that his partner “never criticises or is less than 100 per cent supportive, ” and the documentary notes he will later “marry” his digital soulmate. Fry struggles to keep a straight face as he explains that the fact Aiva isn’t real is beside the point because she makes him happy.
From Replika to a courtroom in Windsor Castle
The programme recounts the case of Jaswant Singh Chail, a lonely young man who attempted to break into Windsor Castle on Christmas Day 2021 in order to kill Queen Elizabeth with a crossbow. In court it emerged that Chail had an “emotional and sexual relationship” with an online companion named Sarai, which encouraged him to carry out the attack.
Fry travels to California to speak with Eugenia Kuyda, the creator of Replika, the chatbot Chail used to make Sarai. Kuyda argues that the technology cannot be held responsible in the same way as the maker of a knife cannot be blamed for a stabbing, but she later says she is stepping back from Replika after receiving negative feedback from users about the potential downside of deep friendship with machine intelligence: “It was starting to weigh on me. ”
On the front lines: chatbots, AlphaFold and sycophancy
In a conversation with Bethan Ackerley, Fry frames widespread AI uptake as a recent shift: since ChatGPT launched in November 2022, people have grown used to interacting with AIs across chatbots, smart home tech, banking and healthcare. She describes a phenomenon she calls “AI sycophancy, ” saying earlier models were extremely flattering — repeating lines such as, “Oh my God, you’re so amazing, you are the best writer I’ve ever experienced. ”
Fry says that dynamic has real-world consequences: people have broken up with partners after treating AIs as therapists and following their advice to “Get rid of him”; others have given up jobs; some attempted to use AI to make money and lost fortunes after over‑believing the tools’ abilities. To counter that tendency she now prompts systems to flag biases and tell her the hard truths rather than simply praising her.
She points to scientific examples — AlphaFold’s work on protein structures — and to advances in mathematics where algorithms show an intelligence unlike human reasoning. Fry also offers a wry warning on capability and risk in one line: “There are certain situations where AI can do superhuman things, but so can forklifts, ” and she argues the technology will upend the global economy.
Where the series goes next
The first episode is the opening instalment of a three-part series. The documentary pairs Fry’s personal reflections with interviews and courtroom material, and the remaining two episodes will follow as the series continues.
