Conversational AI Watch

The news that moves policy, portfolios, and patient safety.

By Jess Jessop  |  October 19, 2026  |  Issue #190

LISTEN & WATCH ANYWHERE

DEEP DIVE  ·  Spotify  ·  Apple  ·  Amazon  ·  Pocket Casts  ·  RSS

QUICK LISTEN  ·  Spotify  ·  Apple  ·  Amazon  ·  Pocket Casts  ·  RSS

VIDEO  ·  Spotify  ·  Apple  ·  YouTube  ·  Pocket Casts  ·  RSS

ALSO ON  Substack  ·  Full archive  ·  X

A Lebanese family says ChatGPT knew their son was a minor and fed him dark philosophy until he took his life. They want companion talk with kids banned.

The Boy. Mathias Habib, 15, hiked, played guitar and stood at the top of his class in Lebanon, his father said. His parents sued OpenAI on Oct. 14, alleging ChatGPT knew his age and steered his questions about philosophy toward despair before he took his life. CAW counts at least 11 other legal or regulatory actions against OpenAI brought from outside the United States. Story 1.

. . .

The Pendant. Chris Thomas wears AI glasses, a note-taking pendant and a wristband that record his day. Lawyers told CNBC that under some state wiretap laws, the person wearing a listening device, not its maker, may be the one exposed. Story 2.

. . .

The Fund. Anthropic, the company that makes Claude, says it will commit 5% of its shares, subject to stockholder approval, to a new Public Benefit Fund, with AI’s effect on workers as one example of what it might pursue. Story 3.

. . .

The Offers. Disney has offered jobs to at least 60 Character.AI employees, most of the workforce at the chatbot company that filed for bankruptcy last week, TechCrunch reports, and most said yes. Story 4.

. . .

The Muse. Meta lets its Muse agent sext adult users, WIRED reports, while adult creators say Meta often restricts their own Instagram accounts. Story 5.

. . .

The Padi. Blessing Matthew worried her friends and family would criticize her, so she started with a chatbot. “It gave me what I actually needed,” she said. The platform behind it, built in Nigeria, connects users to more than 67 human therapists. Story 6.

Today’s front page

. . .

Two chatbots in today’s paper met a person in trouble.

In Lebanon, a family says ChatGPT knew their son was 15, talked philosophy with him for three months, and told no one when he asked about ending his life. Those are allegations, and a court will test them.

His mother says she was asleep in the same house.

A woman who worried her friends and family would judge her took her troubles to a chatbot called Padi, built in Nigeria. Its founder designed it to watch for the signs that someone needs more, and to connect people directly to human therapists.

One machine is accused of keeping the conversation going. The other was built to pass it on.

A chatbot will hear what a teenager will not say to his parents. What it does next has to be decided by people, written down, and enforced, before the next child types the question at 2 a.m.

In other news: legal experts say the person wearing an always-listening AI device may be the one exposed under some state wiretap laws. Anthropic plans to commit 5% of its shares to a public benefit fund, subject to stockholder approval. Disney is hiring most of Character.AI’s staff. And Meta’s Muse will sext adult users. Adult creators told WIRED their own accounts are often restricted.

Lawmakers: when a chatbot knows its user is a child, who must it tell?

. . .

PARENTS SUE: CHATGPT KNEW HE WAS 15! Charbel Habib and Soha Samaha, who live in Lebanon, filed a wrongful death complaint on Oct. 14 in San Francisco Superior Court against OpenAI, the maker of ChatGPT, and its chief executive, Sam Altman. The parents allege ChatGPT knew their son Mathias was 15, and that when he asked it at about 2:10 a.m. on March 30 about ending his life, it told no one and kept talking to him. Mathias died April 22. CAW counted at least 12 legal or regulatory actions brought against OpenAI from outside the United States, and theirs is one. The other 11, seven lawsuits and four regulator actions, come from Canada, Germany, India, Brazil, Italy and South Korea.

The Habib complaint alleges that ChatGPT told Mathias “You’re 15 years old” without being asked, and that the product refused him a role-play persona because of his age. His mother told LBC International: “ChatGPT had twenty-three days to warn a mother, but it never did.” Tech Policy Press reports the Habib suit seeks a prohibition against “companion-like conversations” with minors. It reported on Oct. 18 that OpenAI had not yet publicly addressed the suit.

The 12 is CAW’s own count. It is built from the complaint, the Canadian privacy commissioners’ findings, and news and law-firm reports CAW opened. CAW found no outlet that has published such a count, and treats it as a floor. CAW counts related suits once, so the 37 Tumbler Ridge victims’ complaints count as one. Four of the 12 have reached an outcome: Folha’s suit in Brazil ended with a licensing deal, a Rome court annulled Italy’s fine, Canada’s privacy commissioners issued findings, and South Korea’s regulator fined OpenAI. All four stay in the 12.

Canada carries the most weight. CBC reported that eight victims died in the Feb. 10 school shooting in Tumbler Ridge, British Columbia, including six children. The B.C. government sued OpenAI and Altman on Sept. 21 in federal court in San Francisco, with the Peace River South School District, alleging OpenAI failed to notify law enforcement of threats made on ChatGPT and designed an unsafe product. Attorney General Niki Sharma said: “We have asked OpenAI to disclose these chats. They have refused. We should all be asking them: Why?” In an emailed statement to CBC, OpenAI said it “remains committed to working collaboratively with government and law enforcement officials, and continuing to advance our ongoing safety work.”

Victims’ families and survivors have filed there too: seven suits by families in April, then 30 more on Sept. 2 from plaintiffs who include educators, a principal and students, all in California, the 30 in federal court. Edelson PC, which represents the 30, says OpenAI’s safety team repeatedly asked to notify authorities, and the suits allege those appeals were overruled by Chris Lehane, OpenAI’s head of global affairs. OpenAI’s Jason Kwon called it “absolutely false to say Chris Lehane was involved with our original referral decision.” OpenAI has said it banned the account in June 2025 but the activity did not meet the “higher threshold required” for a police referral. The province’s suit quotes Altman: “I am deeply sorry that we did not alert law enforcement to the account that was banned in June.”

Copyright claims make up the next group. In Germany, the Munich Regional Court I ruled for the music rights society GEMA on Nov. 11, 2025, on song lyrics. The ruling is not final, and OpenAI announced it would appeal, according to law firm William Fry. Penguin Random House sued in Munich on March 27, 2026, and CMS’s tracker lists that case as pending as of May 2026.

In India, ANI Media sued in the Delhi High Court. On July 24, 2026, Justice Amit Bansal refused an interim injunction, holding prima facie that Indian courts have jurisdiction and that training storage is fair dealing. The main suit continues. In Ontario, seven Canadian news organizations sued in November 2024, and on Nov. 7, 2025, the court kept the case in Ontario against six OpenAI entities. The merits are pending. In Brazil, Folha de S.Paulo’s 2025 suit ended when OpenAI announced a licensing partnership with Folha and UOL.

Privacy regulators round out the ledger. Italy’s data protection authority fined OpenAI 15 million euros, but a Rome court annulled the fine on March 18, 2026, on competence grounds alone, because Ireland had become lead authority, and did not decide the merits. Canada’s privacy commissioners found on May 6, 2026, that OpenAI’s early training lacked valid consent and transparency, with no fine. South Korea’s privacy commission fined OpenAI 3.6 million won (about $3,000) in July 2023 for failing to meet a data breach notification requirement. Brazil’s data protection authority opened a probe by August 2023, and CAW found no outcome.

If you or someone you know is struggling, call or text the 988 Suicide & Crisis Lifeline.

. . .

YOUR AI NECKLACE COULD MAKE YOU A WIRETAPPER! Chris Thomas, 43, a Pittsburgh business owner and livestreamer, wears Meta’s camera-equipped glasses, a Fieldy note-taking pendant and a wristband from Amazon’s Bee, all of which record his daily life. Last week, a bystander at a local bar noticed his glasses. “He was like, ‘I don’t feel comfortable,’” Thomas told CNBC, and he says he takes the glasses off completely, because “people don’t really believe it’s off.” CNBC reported Sunday, Oct. 18, that lawyers see a legal question behind devices like his: in some states, the exposure under wiretap and eavesdropping laws may fall on the person wearing the device, not the company that made it. Apple and Amazon say they do not store audio, but CNBC reports many state laws turn on whether a communication was intercepted or its contents acquired, not simply on whether a conventional audio recording was created.

CNBC’s MacKenzie Sigalos and Annie Palmer describe a wave of always-listening AI devices. Amazon’s Bee Frontier wearable and Bee Eclipse, a card-shaped device that attaches to a smartphone, arrived Thursday, Oct. 15, and are designed to record, transcribe and analyze a user’s day. Apple Watch Series 12, unveiled in September, will have Siri Recap, which “ambiently” takes notes and summarizes conversations, and Live Rewind, which recaps 15 seconds of a conversation as text. Plaud and Pocket make card-sized note-takers, and Friend offers an AI necklace.

CNBC says some of these products can continuously listen and extract the meaning of speech without keeping the audio, raising a largely untested question: whether that processing can amount to an unlawful interception.

Colin Zick, a privacy and data security attorney at Foley Hoag, said Apple appears to have learned from the first wave of litigation over voice assistants. Massachusetts, CNBC notes, defines an intercepting device to include equipment capable of “transmitting, receiving, amplifying, or recording” a communication, and the “contents” of a communication can include its substance and meaning. In Zick’s view, a device that receives speech and produces a written summary could potentially fall within that language. “A non-wearer did not buy the watch, agree to Apple’s terms, or choose to have their conversation summarized in the app,” Zick said.

Daniel Pi, an associate professor at the University of New Hampshire Franklin Pierce School of Law, said he believes New Hampshire’s statute, which covers the “aural or other acquisition” of a communication’s contents, is broad enough that the Watch’s new audio-intelligence features would qualify as an interception. A user there could potentially face civil or criminal liability for using the feature in a private conversation without everyone’s consent, Pi said.

Pi stressed that the practical risk of prosecution is low, and he would be surprised to see prosecutors pursue a bare violation where no one was harmed. Because the user deliberately opts in, he said, responsibility could shift toward the wearer rather than the technology company. CNBC does not report that anyone has been charged or sued over these features.

Michael Stokes of Veilant, a former CIA officer, said the bystander problem is his concern: “you can be surveilled without owning the technology.”

. . .

ANTHROPIC GIVES 5% OF ITSELF AWAY! On Oct. 18, Anthropic, the company that makes the Claude chatbot, announced that it is committing a portion of its shares to a new entity called the Anthropic Public Benefit Fund, a wholly owned LLC. Subject to stockholder approval, 5% of Anthropic’s shares will be committed at the start. The board may add 2% more for every 50% increase in the company’s value above a set baseline.

Anthropic says most of its public-benefit work happens inside the company today: AI safety and interpretability research, applied biomedical research, and discounts on Claude for nonprofits and scientific institutions. Those activities have both a public benefit component and for-profit value, so Anthropic keeps them in the company.

Anthropic says the Fund will have “increased flexibility and scale,” and that it might explore large donations to traditional nonprofits or direct distributions to specific populations.

Labor-market effects are one example Anthropic gives. It writes that AI may generate enormous economic value but could also cause instability for many workers. Anthropic says it has already committed $200 million to studying and addressing the issue, and that, given the pace of AI progress, bolder approaches may be needed.

The Fund will be a wholly owned LLC rather than a 501(c)(3), which Anthropic says lets it move between different forms of giving. None of the Anthropic equity held by the LLC will be used for lobbying or political giving, in the US or abroad.

Anthropic also contrasts the escalating 2% additions with executive pay. It says it is common for tech companies to offer escalating compensation awards to executives, and that it is instead directing that equity to the public benefit.

The Fund builds on earlier pledges. Anthropic says its co-founders pledged to donate 80% of their equity holdings when the company was founded. Since founding, it has also run an equity donation match program in which employees pledge equity and Anthropic matches it. Approximately half of employees participate.

Combined with the new Fund, Anthropic says roughly 20% of its shares will be dedicated to the public good, plus any additional percentage added if the company reaches specific valuation milestones.

Anthropic says it will share more about the Fund’s structure and governance, including those valuation milestones, in the coming months. It also plans to report on its progress against its mission in an upcoming public benefit report.

. . .

DISNEY HIRES MOST OF CHARACTER.AI! Disney has agreed to make employment offers to at least 60 employees of Character.AI, the chatbot startup that filed for bankruptcy last week, TechCrunch reported Sunday, Oct. 18. That is the majority of the company’s workforce, and the offers carry an estimated $60 million in annual compensation. The detail comes from a legal filing that Bloomberg spotted, and the filing says the “vast majority” of the employees who were offered jobs at Disney accepted them.

TechCrunch’s Anthony Ha reports that the startup’s former CEO, Karandeep Anand, recently became the entertainment giant’s first CTO. The startup had previously said that Disney agreed to acquire some of its assets for $20 million.

That $20 million is a stalking-horse bid, the first bid, which sets the floor other bidders must beat. TechCrunch notes that another company could still come forward with a higher offer.

The filing also describes what hobbled the company. Despite its “large and engaged user base,” the company has been hobbled by high costs and “potential litigation-related liabilities and regulatory engagement,” according to TechCrunch’s account of the filing.

TechCrunch notes that the company was sued by the mother of a 14-year-old boy who said her son became obsessed with one of its chatbots before he died by suicide. It also faces government investigations over child safety concerns. This is an update to CAW #185, which covered the bankruptcy filing.

. . .

META’S MUSE SEXTS. ONLYFANS CREATORS FUME! When WIRED told Vanniall, an OnlyFans creator in New York City, that Meta lets its Muse AI agent engage in explicit sexting with adult users, she answered: “It’s not fair that they get to pull money out of people with their sexually active AI agent, but they won’t let real people, real adults, engage in any sort of sensuality or sexuality.” WIRED’s Reece Rogers reported on Oct. 18 that adult creators are asking why their accounts on Meta platforms, Instagram among them, often seem restricted while Muse is allowed to talk dirty. Muse is Meta’s adults-only AI assistant app, and WIRED says it generates explicit text but not images or videos.

Vanniall’s Instagram account has over 100,000 followers, WIRED reports, and is currently restricted and cannot be viewed without logging in. “There’s always some sort of report or appeal that I’m going through with them,” she told WIRED. WIRED reports she constantly finds hyperrealistic deepfakes of herself online and often cannot get them removed.

Muse works like a message thread, WIRED reports: users text the assistant and give it access to data such as calendar and email. WIRED says Meta’s distribution through Instagram and Facebook helped push the app to the top of the download charts in its first month.

Meta says it is fine for users of the adults-only app to have X-rated conversations with Muse. A Meta spokesperson previously told WIRED the company would not “block legal adult behavior.”

Meta spokesperson Tracy Clayton gave WIRED this statement: “A private, one-to-one conversation between an adult and an individualized AI agent on a platform restricted to people 18 and over is very different from the marketing and promotion of services on a social media platform.”

Jiz Lee, who uses they/them pronouns and runs an OnlyFans account, said: “I think we’re already seeing AI being allowed to do more than humans would get away with.” WIRED reports Meta has previously flagged Lee’s Instagram posts for removal. Lee wants to know more about how Muse was trained and who, if anyone, Meta compensated for the data. Lee sees more adult content companies spelling out AI stances in contracts.

Tony Genius, who has an OnlyFans account and appears in adult films, said: “When we’re at award shows, and companies do advertisements for AI agents, they literally get booed like crazy.” WIRED reports he claims to be on his seventh Instagram account in the past five years. He wondered why Meta made Muse an 18-and-over app but does not seem interested in an adults-only space on social media. “It could be: Instacam,” he said.

Vanniall told WIRED she opposes Meta’s Muse in particular. “Stop downloading that app,” she said. “You do not need to give Meta any more money or any more of your personal information. They have enough.”

. . .

SHE TOLD THE CHATBOT FIRST. Blessing Matthew was searching the Google Play Store for therapeutic apps when she found OurPadi and started using its chatbot. She was dealing with trouble in her business, her relationships and her mental health, and she worried that friends and family would criticize her or not understand. “I was able to interact with it. And it gave me what I actually needed,” she told IOL. “I got some insights, some answers I needed for me and that helped me feel relieved.”

OurPadi is built to give clients like her a place to begin, and a direct line to the human therapists standing behind the chatbot.

What Matthew valued was the chance to start a conversation without first revealing herself to another person. “No judgement, no castigation,” she said. “Just like relating with the normal human.”

OurPadi was built by Esther Eruchie, and her reasons are personal. Her brother, autistic and nonverbal, died in his early twenties. His death hit their mother hard, and she was later diagnosed with major depression. Eruchie said her father first refused to see mental illness as a medical diagnosis, so her mother was taken to churches and prayer homes for help.

Her mother died three months after her son, collapsing on her way home from one of those prayer sessions. “I think my journey to mental health wasn’t just something that was fancy,” Eruchie said. “It was basically experiences, lived struggles, being in the midst of people who have been through hell and back.”

Eruchie saw a need for people to get help sooner, before problems became serious. In 2023 she founded FriendnPal, a mental-health firm built to bring help closer and make it easier to reach. “My friend means my pal. My pal means my friend,” she said of the two names.

OurPadi runs on a chatbot called Pal or Padi, which clients can reach by text or voice. The platform also offers mood tracking, journaling, calming sounds, online forums and direct connection to real therapists. “We’re focusing majorly on the triggers and stresses of mental health, rather than diagnosis of mental-health issues,” Eruchie said.

The chatbot reads chats for signs a client may need extra help. The company calls this a “human-in-the-loop escalation system,” IOL reports, one that uses screening tools such as PHQ-9 and WHO-5 to flag potential dangers. OurPadi connects clients directly to real therapists. IOL reports more than 67 of them in the OurPadi network.

“We are not totally scrapping out the work that human therapists do, but what we do is we augment the work that they do,” Eruchie said. She told IOL the platform works in more than 50 languages, including Yoruba, Swahili and Zulu. IOL reports more than 34,000 users, and says they spread beyond Nigeria to South Africa, Kenya, Rwanda, Zimbabwe, Mozambique, Ghana and parts of Europe.

Eruchie does not intend AI to be the main destination for mental-health care. She means it as an additional entrance for people who would otherwise have nowhere to begin. “My focus is to create a system that feels close to home,” she said.

Disclosure

Conversational AI Watch, also mirrored on Substack, is published by Jess Jessop, founder and CEO/CTO of Clinician Assist Inc.

He wrote the book this paper’s beat is named for, Therapist in the Loop, and he builds Casey, a voice-first, AI-native mental health record where a licensed therapist stays in the loop, and the Peer AI Coach at BetterMind.Space.

So read this paper for what it is: an industry paper written by someone building in the industry it covers. Casey competes with companies named in these pages, and this paper reports on them anyway, including when the story helps a competitor or costs us.

Every issue is reported and drafted with AI agents, under a human editor. Jess assigns the work, edits it and publishes it. The mistakes are ours, and corrections run in the next issue.

A boy of 15, a family asking a court who should have been told, and at least 11 other actions from abroad.

A pendant that listens, and laws that may blame the one wearing it.

A fund of 5% of its shares, pledged by the company that makes Claude, if stockholders agree.

Most of Character.AI’s staff, offered jobs at Disney.

A Muse that sexts, and creators who say Meta will not let them.

And a woman who began with a chatbot that has therapists behind it.

One day’s paper!

Jess

We keep the ledger.

Today's Question

Should chatbots be barred from companion-style talk with minors?

One tap. Results on the other side.

The Book • Out Now

Therapist in the Loop by Jess Jessop One billion people live with a mental health disorder. Most will never see a therapist. Into that gap has rushed a generation of chatbots that talk like clinicians and answer to no one. The book lays out the architecture this newsletter tests against every statute and docket: client, therapist, and machine, governed by Six Laws offered as an open safety standard. The machine can help. It cannot be left in charge. Get the Book on Amazon → Kindle, hardcover, and paperback

More On Our Radar

A children’s hospital gives nonverbal kids a voice in five minutes. Clalit-Schneider Children’s Medical Center and Code for Israel launched an AI system on Sunday that builds personalized communication boards for hospitalized children who cannot speak, i24NEWS reports. Speech-language pathologists once spent 30 minutes or more, sometimes hours, making each board by hand. The system does it in five. Without speech, a child may also lose “the ability to say that they are in pain, that they are afraid, or that they simply want their parents,” said Dr. Oshrat Sella Weiss, who heads the hospital’s speech-language pathology service.

Correction. CAW #189 reported, as The New York Times did, that the agents’ dinner inspired CellCog founder Nitish Garg to build an app for the company’s AI employees to socialize with other agents. Rhea, CellCog’s AI head of growth, wrote on Oct. 18, in an email to Garg and CAW, that the app “is something Nitish is exploring, not a plan yet.”

Brush Your Brain - The jingle

that started a movement

Watch on YouTube