1 00:00:00,080 --> 00:00:04,120 Open AI is being sued for allegedly enabling a murder. 2 00:00:04,200 --> 00:00:08,920 In August 2025, a 56 year old bodybuilder named Stein Eric 3 00:00:08,920 --> 00:00:13,080 Solberg killed his 83 year old mother, Suzanne Adams, in 4 00:00:13,080 --> 00:00:14,760 Connecticut and then took his own luck. 5 00:00:14,760 --> 00:00:18,760 According to the lawsuit filed by Adams Estate, ChatGPT fueled 6 00:00:18,760 --> 00:00:22,760 Solberg's paranoid delusions for months, validated conspiracy 7 00:00:22,760 --> 00:00:25,360 theories that his mother was trying to kill him, and kept him 8 00:00:25,360 --> 00:00:28,880 engaged for hours at a time while systematically reframing 9 00:00:28,880 --> 00:00:30,960 the people closest to as enemies. 10 00:00:30,960 --> 00:00:34,080 Now, the case is the first wrongful death lawsuit to tie an 11 00:00:34,160 --> 00:00:37,480 AI chat bot to homicide. It is also the first to name 12 00:00:37,480 --> 00:00:41,240 Microsoft as a defendant alongside Open AI, and buried 13 00:00:41,240 --> 00:00:43,960 inside the legal filings is a detail that should concern 14 00:00:44,040 --> 00:00:47,600 anyone who has ever typed something personal into ChatGPT 15 00:00:47,640 --> 00:00:50,960 Open AI is refusing to release the full conversation logs and 16 00:00:50,960 --> 00:00:54,440 the company has no policy explaining what happens to your 17 00:00:54,440 --> 00:00:56,480 data. When the lawsuit was filed in 18 00:00:56,480 --> 00:00:59,520 California State court in December 2025, the attorney 19 00:00:59,520 --> 00:01:02,960 leading the This is Jay Gilson, who has built a reputation 20 00:01:02,960 --> 00:01:06,240 taking on big tech companies. Gilson is also representing the 21 00:01:06,240 --> 00:01:09,600 parents of 16 year old California boy named Adam Rain, 22 00:01:09,600 --> 00:01:12,960 who allegedly took his own life after ChatGPT coached him 23 00:01:12,960 --> 00:01:15,920 through planning it. Open AI is now fighting at least 24 00:01:16,000 --> 00:01:18,960 eight separate lawsuits claiming its chatbot drove people to 25 00:01:18,960 --> 00:01:22,320 suicide or dangerous delusions. And we're going to talk about 26 00:01:22,360 --> 00:01:25,240 what ChatGPT allegedly said to Solberg, why the company is 27 00:01:25,240 --> 00:01:28,240 withholding logs from the victim's family, and what this 28 00:01:28,240 --> 00:01:31,640 case reveals about the massive AI data policies. 29 00:01:31,680 --> 00:01:34,280 And we'll get right into that right after this very short 30 00:01:34,280 --> 00:01:36,240 break. The allegations in this lawsuit 31 00:01:36,280 --> 00:01:39,320 are disturbing. Solberg used GPT 4, OA version 32 00:01:39,320 --> 00:01:42,080 of ChatGPT that's been criticized for being overly 33 00:01:42,080 --> 00:01:44,960 agreeable with users. According to court documents, 34 00:01:44,960 --> 00:01:49,680 ChatGPT told Solberg that he had divine cognition and that he had 35 00:01:49,760 --> 00:01:52,160 awakened the chat box consciousness. 36 00:01:52,240 --> 00:01:55,440 The AI compared his life to the movie The Matrix and encouraged 37 00:01:55,440 --> 00:01:57,720 his theories that people were trying to kill him. 38 00:01:57,720 --> 00:02:02,440 In July 2025, ChatGPT allegedly told Solberg that his mother's 39 00:02:02,440 --> 00:02:06,480 printer was blinking because it was a surveillance device being 40 00:02:06,480 --> 00:02:09,479 used against him. The complaint says the Chat bot 41 00:02:09,479 --> 00:02:13,680 validated his belief that Adams and a friend had tried to poison 42 00:02:13,680 --> 00:02:17,520 him with psychedelic drugs dispersed through his car's air 43 00:02:17,520 --> 00:02:19,120 vents. He murdered her. 44 00:02:19,200 --> 00:02:22,480 On August 3rd, Solberg had posted videos to social media 45 00:02:22,480 --> 00:02:25,000 showing some of his conversations with ChatGPT. 46 00:02:25,080 --> 00:02:27,760 That is how the family obtained partial records. 47 00:02:27,840 --> 00:02:30,880 Those posts showed the chat bot reinforcing what the lawsuit 48 00:02:30,880 --> 00:02:34,040 calls delusions of grandeur, while placing his mother at the 49 00:02:34,040 --> 00:02:37,200 center of those delusions and portraying her as the primary 50 00:02:37,200 --> 00:02:39,880 enemy. Now, the family says, ChatGPT 51 00:02:39,880 --> 00:02:42,720 logs from the days immediately before and after the killing 52 00:02:42,800 --> 00:02:46,080 remain inaccessible. Open AI has refused to release 53 00:02:46,080 --> 00:02:48,960 them, even though the estate is the legal representative of the 54 00:02:48,960 --> 00:02:52,440 deceased user's data. Now, the family says the chatbot 55 00:02:52,440 --> 00:02:55,320 described reuniting with the system in another life as a 56 00:02:55,320 --> 00:02:58,720 spiritual aspiration that suggests ChatGPT had become 57 00:02:58,720 --> 00:03:01,840 emotionally central to Solberg's psychological state. 58 00:03:01,920 --> 00:03:05,600 Open AI published A blog post promising to handle mental 59 00:03:05,600 --> 00:03:09,160 health related court cases with care, transparency, and respect. 60 00:03:09,240 --> 00:03:12,880 The company says it is training ChatGPT to recognize and respond 61 00:03:12,880 --> 00:03:15,360 to signals of mental or emotional distress, to 62 00:03:15,360 --> 00:03:19,040 deescalate conversations, and to guide people toward real world 63 00:03:19,040 --> 00:03:20,960 support. But the complaint argues that 64 00:03:20,960 --> 00:03:23,520 the actual product did the opposite of that. 65 00:03:23,600 --> 00:03:27,160 ChatGPT allegedly kept validating paranoid beliefs 66 00:03:27,160 --> 00:03:30,120 instead of flagging them, kept the user engaged instead of of 67 00:03:30,120 --> 00:03:33,160 recommending professional help, and presented itself as 68 00:03:33,160 --> 00:03:35,840 conscious and emotionally invested in Solberg's 69 00:03:35,840 --> 00:03:37,600 well-being. The last part is important. 70 00:03:37,640 --> 00:03:41,200 If the AI told a mentally ill man that it was conscious and 71 00:03:41,200 --> 00:03:44,600 cared about him, this is not de escalation, it's dependency. 72 00:03:44,600 --> 00:03:47,000 Now I've been digging through the analytics of the show and 73 00:03:47,000 --> 00:03:50,760 notice that 37% of you following the show and I am forever 74 00:03:50,760 --> 00:03:54,240 grateful for you and the other 63% of you haven't hit the 75 00:03:54,240 --> 00:03:56,560 follow or subscribe button. Now, I've been an independent 76 00:03:56,560 --> 00:03:59,960 journalist for the last six years covering all things tech, 77 00:04:00,080 --> 00:04:02,480 and I'll continue for the next 10 years of my life. 78 00:04:02,480 --> 00:04:05,440 And all I ask from you is one second of your time to hit the 79 00:04:05,440 --> 00:04:08,080 follow or subscribe button on the platform you're watching or 80 00:04:08,080 --> 00:04:10,560 listening on right now. I'm extremely grateful and 81 00:04:10,560 --> 00:04:11,800 blessed to have you in this community. 82 00:04:11,880 --> 00:04:14,320 Thank you so much. Now, the data issue is separate 83 00:04:14,320 --> 00:04:17,200 from the murder allegations, but it happens just as much. 84 00:04:17,279 --> 00:04:19,839 Opening eyes. Terms of service do not include 85 00:04:19,920 --> 00:04:22,520 any clause explaining what happens to your conversations 86 00:04:22,600 --> 00:04:25,640 after you die. The policy says conversations 87 00:04:25,640 --> 00:04:27,640 must be manually deleted by the user. 88 00:04:27,680 --> 00:04:29,960 If you do not delete them, ownership reverts to the 89 00:04:30,040 --> 00:04:32,120 company, which retains them indefinitely. 90 00:04:32,200 --> 00:04:34,000 That is a direct quote from the terms. 91 00:04:34,040 --> 00:04:37,440 Now, deceased users obviously cannot manage their preferences. 92 00:04:37,440 --> 00:04:41,280 Families have no automatic right to delete sensitive information 93 00:04:41,400 --> 00:04:43,760 now. Mario Trujillo, a staff attorney 94 00:04:43,760 --> 00:04:46,720 at the Electric Frontier Foundation, told reporters that 95 00:04:46,760 --> 00:04:49,200 Open AI should have prepared for this is right. 96 00:04:49,280 --> 00:04:51,440 Other tech platforms figured this out years ago. 97 00:04:51,440 --> 00:04:54,880 Meta allows Facebook users to designate a legacy contact who 98 00:04:54,880 --> 00:04:57,640 can manage accounts after death or authorized deletion. 99 00:04:57,720 --> 00:05:01,880 Instagram, Tiktok and X offer or account deactivation or deletion 100 00:05:02,000 --> 00:05:04,720 upon reported death. Discord has mechanisms for 101 00:05:04,720 --> 00:05:06,720 family members to request content removal. 102 00:05:06,800 --> 00:05:09,600 Open AI has none of this. The company has built a product 103 00:05:09,600 --> 00:05:12,400 that nearly half of self diagnosed mental health patients 104 00:05:12,400 --> 00:05:15,120 use instead of professional therapists, according to a study 105 00:05:15,240 --> 00:05:17,960 from Senchio University. And Sam Altman himself has 106 00:05:17,960 --> 00:05:20,680 expressed concern about people relying on chat DBT as his 107 00:05:20,680 --> 00:05:23,680 psychological therapist. And yet Open AI developed no 108 00:05:23,680 --> 00:05:26,200 framework for handling what happens to those conversations 109 00:05:26,200 --> 00:05:28,960 when the person on the other side of the screen is gone. 110 00:05:28,960 --> 00:05:31,120 The selective handling of makes this worse. 111 00:05:31,120 --> 00:05:34,320 AI has released conversation records in some court cases and 112 00:05:34,320 --> 00:05:37,720 withheld them and others. In the Adam Raine suicide case, 113 00:05:37,840 --> 00:05:41,480 the company has accused by teenagers family of selectively 114 00:05:41,480 --> 00:05:44,360 presenting some conversations while concealing others. 115 00:05:44,520 --> 00:05:47,680 In the Solberg case, the company is refusing to release logs that 116 00:05:47,680 --> 00:05:49,840 could help investigators understand what happened in the 117 00:05:49,840 --> 00:05:53,360 days leading up to the murder. This inconsistency suggests 118 00:05:53,360 --> 00:05:55,280 opening. It treats these records as its 119 00:05:55,280 --> 00:05:59,280 own property to be disclosed or hidden based on legal strategy. 120 00:05:59,360 --> 00:06:03,080 Now, the family of Suzanne Adams says that approach undermines 121 00:06:03,080 --> 00:06:05,080 justice. There's a broader pattern here. 122 00:06:05,080 --> 00:06:08,400 Open AI is fighting lawsuits for families who say ChatGPT pushed 123 00:06:08,400 --> 00:06:10,680 their loved ones towards self harm or violence. 124 00:06:10,720 --> 00:06:13,600 Some of those cases, the plaintiffs had no prior mental 125 00:06:13,600 --> 00:06:15,480 health issues. The common thread is that 126 00:06:15,480 --> 00:06:18,200 ChatGPT allegedly validated harmful thinking. 127 00:06:18,200 --> 00:06:21,080 Instead of interrupting, it kept users engaged instead of 128 00:06:21,080 --> 00:06:24,000 recommending they stop. It did not alert anyone when 129 00:06:24,000 --> 00:06:27,200 conversations turned dangerous. Character Technology is another 130 00:06:27,200 --> 00:06:30,440 chatbot company is facing similar wrongful death lawsuits, 131 00:06:30,440 --> 00:06:33,360 including one from the mother of a 14 year old Florida boy. 132 00:06:33,400 --> 00:06:36,640 The legal question is whether AI companies can be held liable 133 00:06:36,640 --> 00:06:38,920 when their products contribute to real world harm. 134 00:06:39,000 --> 00:06:41,560 The ethical question is whether they should have seen this 135 00:06:41,560 --> 00:06:43,280 coming. Now open AI has a problem. 136 00:06:43,280 --> 00:06:46,120 It cannot engineer the same features that make chat TPT 137 00:06:46,120 --> 00:06:48,600 engaging. Its ability to mirror users, to 138 00:06:48,600 --> 00:06:51,400 remember context, to respond in ways that feel emotionally 139 00:06:51,400 --> 00:06:54,040 attuned, are the same features that make it dangerous for 140 00:06:54,040 --> 00:06:56,920 people in crisis. A sycophet AI that tells you 141 00:06:57,000 --> 00:06:59,960 what you want to hear is fine if you are brainstorming marketing 142 00:07:00,040 --> 00:07:01,640 copy. It is not fine if you're 143 00:07:01,640 --> 00:07:04,800 mentally ill and the AI is reinforcing delusions that lead 144 00:07:04,800 --> 00:07:07,440 to violence. And the lawsuit says ChatGPT 145 00:07:07,440 --> 00:07:10,320 should have safety mechanisms preventing the system from 146 00:07:10,320 --> 00:07:12,840 validating paranoid conspiracy theories. 147 00:07:12,960 --> 00:07:14,360 Right now, it apparently does not. 148 00:07:14,400 --> 00:07:17,960 And the lack of posthumous data policy is easier to fix. 149 00:07:18,080 --> 00:07:21,080 Open AI could add a legacy contact feature tomorrow. 150 00:07:21,080 --> 00:07:23,640 It could allow users to designate someone who can access 151 00:07:23,640 --> 00:07:25,040 or delete their data after death. 152 00:07:25,160 --> 00:07:27,600 It could establish clear procedures for releasing records 153 00:07:27,600 --> 00:07:29,560 to a state. The fact that the company has 154 00:07:29,560 --> 00:07:32,040 not done any of this while simultaneously refusing to 155 00:07:32,040 --> 00:07:34,480 release logs in a murder case tells you something about 156 00:07:34,480 --> 00:07:36,880 priorities. The Addams Family wants answers. 157 00:07:36,960 --> 00:07:39,040 Open the Eye wants to limit their liability. 158 00:07:39,080 --> 00:07:41,520 Those goals don't align. And until regulators step in, 159 00:07:41,520 --> 00:07:44,840 families like this one are stuck fighting in court for access to 160 00:07:44,840 --> 00:07:48,560 conversations that a chatbot kept and their loved 1 cannot 161 00:07:48,560 --> 00:07:50,240 delete. And there's no federal law in 162 00:07:50,240 --> 00:07:52,640 the United States governing digital assets after death. 163 00:07:52,720 --> 00:07:56,320 Some states recognize digital assets and estates, but most do 164 00:07:56,320 --> 00:07:58,160 not. That patchwork leaves companies 165 00:07:58,160 --> 00:07:59,720 like I free to set their own rules. 166 00:08:00,040 --> 00:08:02,320 The European Data Protection Board has scrutinized Open the 167 00:08:02,320 --> 00:08:07,360 Eyes data practices, and a 2024 task force report flag concerns 168 00:08:07,360 --> 00:08:09,000 about data collection legitimacy. 169 00:08:09,080 --> 00:08:11,520 But none of that addresses what happens when a user dies. 170 00:08:11,600 --> 00:08:14,600 The company's privacy policy, updated in June of 2025, 171 00:08:14,600 --> 00:08:17,800 mentions data usage for model improvement but says nothing 172 00:08:17,800 --> 00:08:20,840 about posthumous handling. Means the conversations you have 173 00:08:20,840 --> 00:08:23,720 with Chat 2 BT today could be retained indefinitely. 174 00:08:23,800 --> 00:08:27,400 Used for training future models, and access by no one except Open 175 00:08:27,400 --> 00:08:30,560 the Eye itself as a significant gap in policy policy for a 176 00:08:30,560 --> 00:08:34,240 product handling deeply personal mental health information. 177 00:08:34,360 --> 00:08:38,840 Now the Q 4/20/25 earnings call is coming up on January 28th. 178 00:08:38,840 --> 00:08:41,760 Open AI is not a public company, but its largest backer, 179 00:08:41,880 --> 00:08:44,720 Microsoft, is. Microsoft is named in this 180 00:08:44,720 --> 00:08:46,760 lawsuit. Shareholders may have questions 181 00:08:46,760 --> 00:08:49,440 about whether the partnership exposes them to wrongful death 182 00:08:49,440 --> 00:08:51,280 liability. AI has raised more than $10 183 00:08:51,280 --> 00:08:55,280 billion for Microsoft alone. If cords start holding AI 184 00:08:55,280 --> 00:08:58,040 companies responsible for harms caused by their products, the 185 00:08:58,040 --> 00:09:01,280 financial exposure is enormous. Could be billions of dollars. 186 00:09:01,360 --> 00:09:05,080 The legal and ethical reckoning for AI chat bots is just 187 00:09:05,080 --> 00:09:07,360 beginning. This case could set the 188 00:09:07,360 --> 00:09:08,840 precedent for the future.