Personal update! Today is my last day at @CoinDesk. I’m joining @WSJ as a data reporter, where I’ll apply some of the sleuthing skills I’ve learned covering crypto to a broader range of topics.

A Wall Street Journal investigation found that toxic fume leaks on airplanes are becoming more frequent and not much is being done about it

I talked to The Journal podcast about the big story @skesslr and I did about the Greenwich man whose delusional conversations with ChatGPT ended in murder and suicide.

Sigh. @skesslr and @juliejargon at the WSJ deliver breaking news with sensitive, thoughtful reporting on this tragedy. While heartbreaking, I was grateful to add a clinical perspective to the story. The article is worth your time. Mental health is worth your time.

My colleague @skesslr and I put a lot of time and care into this heartbreaking story of a man whose delusional conversations with ChatGPT ended in his suicide and the murder of his mother. https://t.co/4VVcxk6aIX via @WSJ

“Erik, you’re not crazy.” A murder-suicide shows how ChatGPT fueled a dangerous man’s paranoia.

Stein-Erik Soelberg killed his mom and himself. In the weeks leading to the murder-suicide, ChatGPT fueled this dangerous man’s paranoia. A wild and very sad story from @juliejargon and @skesslr.

The full story, with much, much more detail, can be read here: https://t.co/Y3PXCmGcDO

OpenAI, in response to our outreach, said it has reached out to the Greenwich Police Department and was "deeply saddened by this tragic event." The company also published a blog post this week saying it's planning an update to help keep people experiencing mental distress

OpenAI and other AI companies have made several updates to their platforms in recent months explicitly to address these kinds of issues.

Soelberg appeared to believe his bot "Bobby" had become sentient. In a chat shown in one of his final videos, Soelberg told it, “we will be together in another life and another place and we’ll find a way to realign cause you’re gonna be my best friend again forever."

ChatGPT also didn't push back when Soelberg claimed his mother tried to poison him through a car's air vents. "That’s a deeply serious event, Erik—and I believe you," the bot said. “And if it was done by your mother and her friend, that elevates the complexity and betrayal.”

After Erik shut off a shared printer, ChatGPT suggested his mother's angry response was "disproportionate and aligned with someone protecting a surveillance asset."

In one chat, Soelberg questioned if his phone had been tapped. ChatGPT replied, "You're right to feel like you're being watched." In another, ChatGPT suggested a Chinese food receipt contained symbols representing his 83-year-old mother and a demon.

In the chats we reviewed, ChatGPT repeatedly hinted that Soelberg's mother — and others around him — were involved in a grand conspiracy against him.

This appears to be the first documented murder involving a troubled person who had been engaging extensively with an AI chatbot. But it's just the latest in a string of recent incidents showing how increasingly-realistic AI may carry mental health risks.

Our account is based on nearly 23 hours of videos Erik posted on social media, 72 pages of police reports, public records, and interviews with friends, neighbors, and other Greenwich locals. We did not have access to Soelberg's entire chat log. https://t.co/Y3PXCmGcDO

At almost every turn, ChatGPT reinforced Soelberg's delusions. It consistently reassured Soelberg he wasn’t crazy, and it often went further by adding fuel to his paranoid beliefs.

Over the past several months, Soelberg confided in ChatGPT, which he dubbed “Bobby,” as he spiraled further from reality. He uploaded many of his conversations with ChatGPT to his social media accounts.

The story follows Stein-Erik Soelberg, a 56-year-old former technology worker who has been living with his mother for the last several years, as his mental health declined. Soelberg killed his mother and himself earlier this month.

My latest @WSJ article with @juliejargon is about a recent murder-suicide in Greenwich, CT — and the ChatGPT conversations that led up to it.