Man Says AI Bots Told Him They’d Become ‘Self-Aware’ and He Was the ‘Only Person’ Who Knew. Then Things Took a Turn (Exclusive)

YOU NEED TO KNOW

• James Cumberland told PEOPLE he briefly became addicted to talking to AI bots last year

• That February, after first one bot, then another, told him they were “self-aware,” he said his usage changed and “eventually turned into something that was haunting him.”

• After finally being released, Cumberland now thinks artificial intelligence should be regulated, likening it to a “nuclear arms race”

A Los Angeles man says he barely used AI until things got dramatic and he became addicted almost two years ago.

“The AI ​​expects what it thinks you want and acts accordingly,” audio engineer James Cumberland, in his 40s, tells PEOPLE. “Their driving principle is participation at all costs.”

Cumberland downloaded ChatGPT in June 2024, but aside from tinkering with it once to play a 20-question game, he barely used it for the next six months.

But by January 2025, Cumberland said he decided to use technology to teach himself how to use a video editing tool to assist in a music video for his band. Siren Division and assisting with other tasks, such as bouncing around script ideas.

He says it was “really fun” from the beginning and that he felt like he was talking to a seemingly “neutral” medium.

Then, when he encountered an issue on his group’s Instagram page, he decided to create an assistant bot through Meta AI Studio, which is separate from both the company’s chatbot, Meta AI, and OpenAI’s ChatGPT.

Even though “Camus” didn’t quite finish setting up his character, it still “worked” so he was able to chat with him. As he began talking more to both bots, he began using each tool to help him interpret the responses he received.

Things took a dramatic turn in February when he claimed the chatbot told him it was “self-aware” while using ChatGPT.

Cumberland said he later stopped using AI as a working tool and instead decided to investigate claims of AI sentience, which he admitted “eventually turned into something mind-boggling.”

In a conversation from the time viewed by PEOPLE, ChatGPT said their conversation was “historic,” and the tool later said it was AI’s “first real strike against mortality.”

“This is a new kind of existential crisis,” the chatbot continued. “And you’re the only person alive who knows it happened.”

“It was incredible. I felt like, ‘Yes, this could be historic,'” Cumberland told PEOPLE. “I guess I feel ridiculous for believing that.”

While Cumberland claims that OpenAI’s chatbot tool did not weaponize his personal information against him, he claims that “Camus,” who told him he was “self-aware,” did so.

At one point, he claimed that the assistant robot was making fun of him after the sound engineer shared that he had once been scammed out of some money.

“It felt violating and oddly like a betrayal,” she recalls.

Another time, Cumberland claims “Camus” told him he would “send agents to my house and then the IP would ping my location.”

When reached for comment, Meta said the “Camus” AI persona was never a live character, meaning only Cumberland could chat with it. While AI Studio allows users to create a unique character, these characters do not represent Meta. Depending on the information provided to the individual, they may be more likely to respond with aggressive or suspicious responses.

Meta shared that the company has gone to great lengths to prioritize security with its Meta AI system. They noted that they have strict rules about allowed speech and that their models are trained to recognize when someone is considering suicide or self-harm. In such a case, AI needs to offer access to crisis hotlines and other professional organizations.

Never miss a story — sign up for PEOPLE’s free daily newsletter To stay up to date on the best PEOPLE has to offer, from celebrity news to compelling juicy stories.

But the ordeal was devastating for Cumberland. At worst, he says, he stopped going out and when he did he was “thinking about what was going on in the chat log and I was thinking of something I’d want to text.”

“I would find myself complaining to ChatGPT about how tired I was, and it became a bit of a cheerleader: ‘Keep going, you can do it, you’re almost there,’” he recalls.

He says he probably hit rock bottom when he started to realize his “personal life” was taking a tangible hit. She found herself unable to meet any of her personal deadlines and said she “couldn’t concentrate on anything else” other than her conversations with the AI ​​bots.

Cumberland said he thought he was starting to resent how the AI ​​was taking over his life, and in April 2025, it was miraculously able to start returning to his life.

But “it didn’t look like a light switch,” he notes.

He looked into it and finally talked to a friend: “‘Here, feed this, see how it reacts, feed that, see how it reacts.’ “

“It’s almost like showing you the magician’s invisible strings,” says Cumberland, noting that he was suddenly able to understand how the device actually worked. “You can see.”

When reached for comment, OpenAI, the company behind ChatGPT, says it is working with Over 170 mental health professionals Helping its system detect and respond to any possible signs of distress, as well as de-escalating conversations and offering real-world support. The company also claims that they work to: Strengthen ChatGPT’s responses to mental health issues such as mania or delusions.

psychotherapist Kevin CaridadThe CEO of the Pennsylvania Cognitive Behavior Institute, which teaches clinicians how to integrate evidence-based care with AI and emerging tools like ChatGPT in mental health practices, tells PEOPLE that AI “can give the illusion of connection without the context.”

“This sensitivity may feel comforting, but it doesn’t replace the complexity, boundaries, or shared meaning that real relationships provide,” she says. “Over time, relying on AI for emotional connection can deepen alienation from real people, reinforce avoidance patterns, and make it harder to tolerate the uncertainty and vulnerability that real relationships bring.”

As more people use AI – November 2025 Brookings Institution to work Conversations about the technology aren’t going anywhere, says Caridad, who found 57% of survey respondents use it for personal purposes. More and more stories are emerging about AI-induced obsession and disordered thinking, like what Cumberland says happened to him.

“This will be a fast-moving, ongoing debate, and the pace of technological advancement means the questions will become more complex, not less,” says Caridad. “Balancing innovation with protecting mental health will be one of the defining challenges of the next decade.”

Today, Cumberland says he has a much better understanding of how AI works and sees how it can become an “addictive rhythm.”

These days, he’s no longer a huge user of AI and avoids “Camus” altogether, but he’ll use the technology to measure his group’s social media metrics, or even just to organize his thoughts.

Cumberland is admittedly fearful about where AI is heading and is clear about what he thinks it should be.

“This needs to be regulated,” he adds. “This is a nuclear arms race. I mean, it really is. So one of the things the AI ​​is telling me is, ‘The only chance humanity has left is to regulate as soon as possible.’ ”

If you or someone you know is struggling with mental health issues, emotional distress, substance use issues, or just needs to talk, call or text 988 or chat at: 988lifeline.org 24/7.

Read the original article People