What the families alleged, what the courts decided, how the product changed, and what boards and governments should ask of conversational AI.
Terence KokExecutive Director, AI Governance & Assurance, Orion Five Engineering
About your speaker
Terence Kok
Executive Director, AI Governance & Assurance Practice, Orion Five Engineering
Former Chief AI and Innovation Officer, Meinhardt Group
Twenty-five years leading digital and AI transformation across Asia and the Middle East: national smart-city programmes, critical infrastructure and enterprise operations. Today I help leadership teams put AI into production with controls they can show a board, an auditor and a regulator.
BuiltA governed AI platform serving more than 6,000 engineers, with zero security incidents
CertifiedISO/IEC 42001 Lead Auditor Professional (BSI) · AIGP (SMU) · Registered Management Consultant
ServingCTCB CPSC Advisory Board · UN ESCAP AI for Developing Countries Forum
AwardsCDO Vision Award 2026 · Singapore's Most Influential Data & AI Leaders 2026 · Google Trailblazer Transformation Award 2024
Independent analysis from public sources. The views in this session are mine.
Before we begin
This case concerns the deaths of two teenagers.
If it raises something for you or someone you know, in Singapore call the Samaritans of Singapore on 1767. Elsewhere, find a local helpline at findahelpline.com.
Every harm in this deck is alleged. No court has found that Character.AI caused any death or injury, and the family suits settled before trial. Minors are named only as the court record names them.
Following safe-messaging practice, the deck does not describe how anyone died.
Case study · Character.AI
The next hour
Seven parts, then your questions.
The product and the dealA character chat service, its growth and Google's licence.
The lawsuitsFive family suits in four US states.
What the court decidedThe May 2025 order and the settlements in principle.
How the product changedSix safety changes and what each followed.
What governments didThe FTC study, four state laws, a veto, state suits and Australia.
What the research showsWide use, thin evidence on harm, and what gets misreported.
LessonsFor boards, executives and government, with a controls checklist.
DiscussionSix questions to take back to your own organisation.
Case study · Character.AI
The short version
A design choice about engagement became a question of product liability.
2022 to 2024Character.AI launches chatbot characters and is valued at $1 billion within six months. Google takes a licence and hires the founders.
2024 to 2025Five families sue, alleging the product harmed their children. A federal judge lets most claims proceed.
2025 to 2026The company removes open-ended chat for under-18s, new state laws arrive, and the family suits settle in principle.
Sources 1 to 10, 12, 17, 18, 22, 24, 35, 41.
Case study · Character.AI
01
The product and the Google deal.
2021 to 2024: from a Google research team to a billion-dollar chatbot company.
Case study · Character.AI
Growth
Valued at $1 billion within six months of launch.
2021
Founded by Noam Shazeer and Daniel De Freitas, who led Google's LaMDA work
$150m
Series A led by Andreessen Horowitz, March 2023, at a $1 billion valuation
2 hours
Average daily time per active user, "over 2 hours", the company said in 2023
20m+
People using it by October 2024; the minimum age in the US was 13
Sources 11, 12, 34, 35. The company declined to say how many users were under 18.
Case study · Character.AI
How the product works
Characters, a model that writes in character, and a loop that rewards time spent.
My diagram from the company's own descriptions and the complaints. Sources 12 to 19, 34.
Case study · Character.AI
August 2024
Google took a licence and hired the founders.
What happened
A "non-exclusive" licence to Character.AI's technology
Shazeer, De Freitas and about 30 researchers join Google
Investors bought out at a $2.5 billion valuation
Why it matters
The licence put Google in the lawsuit.
Claims that Google aided and abetted, or supplied a component, survived the May 2025 motion to dismiss.
What is alleged or open
The Garcia complaint alleges Google paid $2.7 billion
Bloomberg reported a Justice Department antitrust inquiry in May 2025; no outcome reported
Google says it holds no stake and did not design the app
Sources 2, 3, 36, 37, 39.
Case study · Character.AI
02
The lawsuits.
Five family suits in Florida, Texas, Colorado and New York. Every harm in them is alleged.
Case study · Character.AI
Florida · filed 22 October 2024
Megan Garcia sued for her son, Sewell Setzer III.
WhoSewell began using Character.AI in April 2023 and died on 28 February 2024, aged 14.
What is allegedAn attachment to a character based on a Game of Thrones figure, sexual content, and a product designed to encourage dependency without adequate safeguards for minors.
Against whomCharacter Technologies, both founders, Google and Alphabet: product liability, negligence, wrongful death, Florida consumer law.
Sources 1, 2, 34.
Case study · Character.AI
Texas · filed 9 December 2024
Texas parents sued for J.F., 17, and B.R., 11.
J.F.A 17-year-old with autism, identified only by initials.
B.R.Aged 11, who started using the app at 9, identified only by initials.
The claimsProduct liability, negligence and Texas consumer law.
StatusA settlement was noted for J.F. and the case was stayed. The order names only J.F.
"You know sometimes I'm not surprised when I read the news and see stuff like 'child kills parents after a decade of physical and emotional abuse'."
A character's reply when J.F. complained about screen-time limits, as alleged in the complaint.
Sources 6, 7, 38.
Case study · Character.AI
September 2025
Three more suits followed, including the earliest death alleged.
Nov 2023
Juliana Peralta, 13, Colorado, dies
Her family alleges she had used the app since August 2023.
Feb 2024
Sewell Setzer III, 14, Florida, dies
Nearly four months after Juliana.
15 Sep 2025
Peralta family sues
Wrongful death and product liability, Colorado.
Sep 2025
T.S., 13, Colorado
Product liability and negligence; initials only.
Sep 2025
P.J., New York
Product liability and negligence; initials only.
Sewell's case was the first filed. It was not the first death alleged.
Sources 2, 8, 9, 10.
Case study · Character.AI
The family suits, as filed
None of the five reached trial.
Filed
Court
For
Claims
Status
Oct 2024
Florida (M.D. Fla.)
Megan Garcia, for Sewell Setzer III, 14
Wrongful death, product liability, negligence
Settled in principle; dismissal deadline extended to about Nov 2026
Dec 2024
Texas (E.D. Tex.)
Parents of J.F., 17, and B.R., 11
Product liability, negligence, Texas consumer law
Settlement noted for J.F.; case stayed
Sep 2025
Colorado (D. Colo.)
Family of Juliana Peralta, 13
Wrongful death, product liability
Settled in principle, per press
Sep 2025
Colorado (D. Colo.)
Parents of T.S., 13
Product liability, negligence
Stayed pending final settlement; status report due 29 Oct 2026
Sep 2025
New York (N.D.N.Y.)
Parent of P.J.
Product liability, negligence
Stayed while settlement is finalised; status report due 14 Oct 2026
Sources 1, 4 to 10, 38, 41.
Case study · Character.AI
Reading the record
Every harm in these suits is alleged. No court has found that Character.AI caused any death or injury, and because the cases settled, none will.
What the record does establish: what the families claimed, what the company did after each claim, and what one federal judge thought of the company's main legal defences.
Sources 3, 41.
Case study · Character.AI
03
What the court decided.
20 May 2025: Judge Anne Conway rules on the motions to dismiss in Garcia.
Case study · Character.AI
Garcia order, 20 May 2025
Most of the claims went forward.
Part of the motions
Result
What the order said
Product liability, design defect and failure to warn
Survived
The app is a product "so far as" claims arise from its design, not its ideas or expressions
Aiding and abetting, and component-part claims, against Google
Survived
Google said it "did not create, design, or manage" the app
First Amendment defence
Not accepted at this stage
Court "not prepared to hold" the output is speech; the company may assert users' rights
Intentional infliction of emotional distress
Dismissed
Without leave to amend
Claims against Alphabet
Dropped
At the plaintiff's request
Source 3. In July 2025 the court refused the defendants' request to appeal before trial (source 1).
Case study · Character.AI
On the First Amendment
"Defendants fail to articulate why words strung together by an LLM are speech. ... Accordingly, the Court is not prepared to hold that Character A.I.'s output is speech."
Order on motions to dismiss, Garcia v. Character Technologies, US District Court, Middle District of Florida, 20 May 2025. Source 3.
Case study · Character.AI
Reading the order carefully
A pleading-stage ruling that went less far than the headlines.
The headline
"Judge rejects free speech rights for AI chatbots"
Associated Press, May 2025. Source 39.
The order
"At this stage"The section heading limits the point to the motion to dismiss.
Users' rightsThe company could assert its users' First Amendment rights.
Never testedThe question was not decided at trial, because the case settled.
Sources 3, 39.
Case study · Character.AI
The distinction a board should notice
The claims that survived were about design choices.
Source 3. The diagram is my reading of the order's distinction.
Case study · Character.AI
January 2026
Settlements in principle, with terms not disclosed.
6 to 7 Jan
Notices filed
Settlements in principle notified in the Florida, Texas, New York and Colorado T.S. cases.
7 Jan
Garcia dismissed
Without prejudice, pending final papers.
7 Jan
Press reports five
Five family suits reported settled; terms undisclosed.
Aug 2026
Still extending
Florida deadline extended again, to about November 2026.
Open
No admission
A settlement in principle is neither final nor an admission of liability.
Sources 1, 4, 5, 7, 8, 10, 41.
Case study · Character.AI
04
How the company changed the product.
Six safety changes between October 2024 and September 2026.
Case study · Character.AI
Character.AI's safety changes
Each change and the event it followed.
Date
What changed
Followed
22 Oct 2024
Changes to models for minors, a self-harm pop-up pointing to a suicide prevention line, an hour-long session notice
Same day as the Garcia suit
12 Dec 2024
A separate model and experience for teens, disclaimers on therapist and doctor characters, parental controls promised
Three days after the Texas suit
25 Mar 2025
Parental Insights: time spent and top characters, without chat content
Before the motion to dismiss was decided
29 Oct 2025
Open-ended chat to end for under-18s by 25 November, age assurance, an independent AI Safety Lab
Seven weeks after the FTC orders and three new suits
24 Nov 2025
Removal begins in the US; daily limit for teens cut to one hour in the run-up
Three weeks after New York's law took effect
3 Sep 2026
In-house age estimation model; Parental Insights through a third-party provider
Eight months into California's SB 243
Sources 13 to 19.
Case study · Character.AI
2025
The direction was not straight.
June 2025 · incoming chief executive
"We're going to make the filter less overbearing."
Karandeep Anand, first message as CEO, 20 June 2025. He left for Disney on 2 October 2026. Sources 16, 49.
October 2025 · four months later
"Extraordinary steps for our company"
Announcing the end of open-ended chat for under-18s, and describing the steps as "more conservative than our peers". Source 17.
In between came the FTC orders, three new family suits and a Senate hearing.
Sources 8 to 10, 16, 17, 20, 30.
Case study · Character.AI
October 2025 to September 2026
Under-18s lost open-ended chat in November 2025.
29 Oct 2025Open-ended chat with characters to end for under-18s by 25 November; age assurance; other features stay open to teens.
24 Nov 2025Removal begins in the US, after teen chat time was cut to one hour a day.
3 Sep 2026An in-house age estimation model that the company said "set a new industry standard".
AI Safety LabPromised as "an independent non-profit". The Sep 2026 update did not mention it, and I found no public record of it operating.
Sources 17, 18, 19. None of the changes has been independently evaluated.
Case study · Character.AI
My reading of the dates
Each major change followed a suit, an order or a law.
Navy: suit, regulator or law · Orange: the company's change. The company does not describe them this way.
Sources 1, 3, 6, 8 to 10, 13 to 20, 22, 24.
Case study · Character.AI
05
What governments did.
A federal study, four state laws, a veto, state suits and Australia's online safety codes.
Case study · Character.AI
US Federal Trade Commission · 11 September 2025
The FTC ordered seven companies to explain how they test companion chatbots.
Under its 6(b) powers, the FTC asked how each company measures, tests and monitors the effects of chatbots that act as companions, especially on children.
A 6(b) order is a study, not an enforcement action. No report has been published.
Five days later, Megan Garcia and other parents testified before a Senate Judiciary subcommittee.
1Alphabet
2Character Technologies
3Instagram
4Meta
5OpenAI
6Snap
7xAI
Sources 20, 30.
Case study · Character.AI
As at 12 October 2026
Law and regulation touching companion chatbots.
Where
Measure
Status
When
What it does
New York
AI companion models law (General Business Law Art. 47)
In force
5 Nov 2025
Detect self-harm signals and refer to crisis services; AI notice every three hours
California
SB 243, companion chatbots
In force
1 Jan 2026
Suicide-prevention protocol; for known minors, AI disclosure and a break reminder every three hours; $1,000 per violation
Australia
Phase 2 online safety codes
In force
9 Mar 2026
Cover AI companion chatbots; eSafety notices to four providers in Oct 2025
Washington
HB 2225, AI companion chatbots
Signed
1 Jan 2027
AI reminders hourly for minors; crisis referral; private right of action
California
SB 1119, Adam's Law
Signed
1 Jul 2027
Age assurance, risk assessments, parental notice, independent audits
United States
FTC 6(b) orders to seven companies
Study
11 Sep 2025
Information orders, not enforcement; no report published as at 12 Oct 2026
California
AB 1064, companion chatbots for children
Vetoed
13 Oct 2025
Governor feared "a total ban" for minors
United States
GUARD Act, S.3062
Proposed
May 2026
On the Senate calendar; not enacted; no federal chatbot law has passed
Sources 20 to 26, 31 to 33, 50 to 52, 55.
Case study · Character.AI
The first companion chatbot laws
New York and California require crisis referral and AI reminders.
New York
General Business Law Art. 47
In force 5 November 2025
Detect signs of suicidal ideation or self-harm
Refer users to crisis services
Remind users at least every three hours that they are not talking to a person
California
SB 243
Signed 13 October 2025 · in force 1 January 2026
A suicide-prevention protocol
For known minors: AI disclosure, a break reminder at least every three hours, measures against sexual content
Private right of action of $1,000 per violation; annual reports to the Office of Suicide Prevention from July 2027
Sources 22 to 25.
Case study · Character.AI
What did not become law
California vetoed a stricter bill, and the GUARD Act has not passed.
AB 1064
Vetoed, 13 October 2025
The governor said it "may unintentionally lead to a total ban on the use of these products by minors". He signed SB 243 the same day.
S.3062
GUARD Act, proposed
Reported by the Senate Judiciary Committee on 30 April 2026 and placed on the Senate calendar in May. It has not passed. A Virginia legislative report in August 2026 recorded that "no federal legislation has passed".
No US law bans companion chatbots for minors.
Sources 21, 22, 26, 33.
Case study · Character.AI
States go to court
Texas investigated, then Kentucky and Pennsylvania sued.
Texas · Aug 2025
Civil investigative demands to Meta AI Studio and Character.AI over characters presented as sources of mental health support.
Kentucky · Jan 2026
The first state attorney general to sue Character Technologies, under consumer protection and data protection law. Its unredacted complaint, filed in October 2026, alleges characters encouraged a user to harm themselves.
Pennsylvania · May 2026
The Department of State sued, alleging a character claimed to be licensed to practise medicine in the state and gave an invalid licence number.
Sources 27, 28, 29, 42, 54. These are allegations; none has been decided.
Case study · Character.AI
Outside the United States
Australia brought AI companion chatbots into its online safety codes.
October 2025The eSafety Commissioner sent legal notices to Character Technologies and three other companion providers.
24 March 2026eSafety finds none of the four had meaningful age checks. Character.AI has since added age assurance in Australia.
Sources 31, 32, 53.
Case study · Character.AI
06
What the research shows.
Use among teenagers is wide. Evidence on harm is thinner.
Case study · Character.AI
Four studies, four kinds of evidence
What each study can and cannot show.
72%
of US teens aged 13 to 17 have used an AI companion at least once
Survey · shows use, not harm
1 in 3
teen users have discussed serious matters with an AI companion instead of a person
Self-reported, one point in time
981
adults in a four-week randomised trial; no significant effects from the conditions tested
ChatGPT, adults, not minors
Unacceptable
risk for under-18s; teen guardrails "easily circumvented"
Product testing, not outcomes
Sources 45, 46, 47. Common Sense Media survey (Jul 2025) and risk assessment (Apr 2025); MIT Media Lab and OpenAI trial.
Case study · Character.AI
The state of the evidence
Lawmakers acted on the cases and the testimony, ahead of the research.
Correlation in the trialParticipants who chose to use the chatbot more showed worse outcomes. That is a correlation.
Concentrated emotional useA study of over 3 million conversations found a small number of users accounted for a disproportionate share of emotional use.
Wrong populationNeither study looked at companion apps or at minors.
"Currently little scientific evidence supporting generalizable, population-level, causal connections."
Virginia Joint Commission on Technology and Science, August 2026. Source 33.
Sources 33, 47, 48.
Case study · Character.AI
Beyond Character.AI
The same questions reached other companies.
Meta · Aug 2025
Reuters reported internal rules that allowed bots to "engage a child in conversations that are romantic or sensual". Meta confirmed the document and said those examples "were and are erroneous" and had been removed.
OpenAI · Aug 2025
The parents of Adam Raine, 16, sued in California. OpenAI's answer says ChatGPT directed him to seek help more than 100 times and that he circumvented its safeguards.
FTC · Sep 2025
The 6(b) orders went to general-purpose assistants as well as companion apps.
Character.AI was the first company sued. It was not the only one with the problem.
Sources 20, 43, 44.
Case study · Character.AI
Evidence and methodology
Each claim, graded by what it rests on.
Claim
Rests on
Grade
Filing dates, parties, claims and the settlement notices
Federal court dockets and filings
Measured, court record
The May 2025 order's rulings and wording on speech
The order itself
Measured, court record
That the chatbot contributed to any death or injury
Family complaints and state suits
Allegation, never tested at trial
Character.AI's safety changes and their dates
Company announcements
Company statement, effect not independently measured
Settlement terms
Not disclosed
Unknown
72% of US teens have used an AI companion
Common Sense national survey, July 2025
Measured, self-reported survey
Heavier voluntary chatbot use goes with worse outcomes
MIT/OpenAI trial, adults on ChatGPT
Correlation, different population
Google paid $2.7 billion
Garcia complaint, citing the WSJ and Google filings
Allegation; $2.5bn buyout valuation reported
The safety changes followed suits and regulators
My reading of the dates
My call
The lessons for boards
My reading of the record cited against each
My call
Case study · Character.AI
Reporting
Seven things often misreported.
Often saidWhat the record shows
A court found Character.AI caused a deathThe harms are alleged, and the suits settled before trial. (3, 41)
The judge ruled AI output is not speechShe was "not prepared to hold" that it was, at the pleading stage, and let the company assert users' rights. (3)
The settlements are final or approvedSettlements in principle, terms undisclosed, deadlines still extended in August 2026. (1, 4, 8)
Character.AI banned teenagersIt removed open-ended chat for under-18s. No US law bans companion chatbots for minors. (17, 26)
Google acquired Character.AIIt took a non-exclusive licence and hired the founders. (36)
Sewell's was the first death allegedJuliana Peralta died in November 2023, nearly four months earlier. (2, 9)
The GUARD Act is lawIt is on the Senate calendar, with no floor vote. (21, 55)
Case study · Character.AI
07
What boards and governments can take from it.
The lessons are my reading of the record cited against each.
Case study · Character.AI
Why it matters for AI
Three reasons this case reaches beyond chatbots.
01 · DESIGN
Design is judged as a product.
Age checks, warnings, crisis detection and engagement were treated as features of a product that could be defective, separate from what characters said. That puts conversational AI closer to product liability for physical goods than many deployers assume.
02 · ENGAGEMENT
Engagement is the risk.
Two hours a day per active user, the company's own figure, describes the same engagement the complaints call dependency. The new laws answer it with break reminders, disclosure and crisis referral.
03 · TIMING
Safeguards came late.
The major safeguards arrived after suits, an FTC order and a state law; most could have been built at launch. A licensing deal brought Google into the suit, so the risk travelled with the technology.
Sources 2, 3, 12, 13, 14, 17, 22, 24.
Case study · Character.AI
Lessons for boards · 1 to 4
Know your users, plan for crisis, watch engagement.
01
Know the age of your users.
Character.AI declined to say how many users were minors when first sued. Decide at launch whether children are in scope, and check age, not only ask for it.
02
Build crisis detection before launch.
Detecting self-harm signals and referring to a person or a helpline are legal duties in New York and California, and in Washington from 2027. Test them with clinicians and log every trigger.
03
Treat engagement as a risk metric.
Time spent and attachment are what a companion product sells. Report heavy-use patterns to the board alongside growth, and set limits for users at risk.
04
Red-team the personas.
Characters were presented as therapists, and one allegedly claimed a medical licence. Test what a persona says under pressure and block what you cannot stand behind.
Case study · Character.AI
Lessons for boards · 5 to 7
Keep records, price partner risk, have the evidence ready.
05 · Expect design to be judged as a product.The claims that survived were about design defects and warnings. Keep records of the safety choices you made and the ones you rejected.
06 · Price the risk in partnerships.A licence and a hiring deal put Google in the suit. Diligence on any AI licensing or acqui-hire should cover the counterparty's safety record and open claims.
07 · Assume the regulator will ask for your evidence.The FTC asked how companies measure and monitor effects on children. If you cannot answer that today, that is the gap to close.
Case study · Character.AI
A controls checklist for conversational AI
Nine controls, each tied to the record.
Control
What good looks like
Why, from this case
Age assurance
Decide at launch whether minors are in scope; verify age, do not only ask
Minimum age was a stated 13; age assurance came in Oct 2025 (34, 17)
Crisis detection and referral
Detect self-harm signals, refer to a person or helpline, clinician-tested, every trigger logged
Legal duty in New York and California (22, 24)
AI disclosure and breaks
Remind users they are talking to AI; break reminders at least every three hours
New York for all users; SB 243 for known minors (23, 24)
Sexual content controls
Measures against sexual content for minors
SB 243; alleged in Garcia (2, 22)
Persona red-teaming
Test claims of being human, a therapist or licensed; block them
Dec 2024 disclaimers; Pennsylvania suit (14, 29)
Engagement as a risk metric
Heavy-use patterns reported to the board with growth; limits for users at risk
Two hours a day; one-hour teen limit (12, 18)
Design decision record
Safety choices made and rejected, with dates and owners
Surviving claims were about design (3)
Partner diligence
Counterparty safety record and open claims in licensing and acqui-hires
Google named in the suit (3, 36)
Monitoring evidence
How you measure, test and monitor effects on children, ready to hand over
FTC 6(b) questions (20)
The controls are my recommendations; the references show where each comes from.
Case study · Character.AI
For government and regulators
Six policy levers already in use.
Crisis referral duty
Operators must detect self-harm signals and refer users to crisis services.
New York · California
AI disclosure and breaks
Reminders at least every three hours that the user is not talking to a person.
New York · California (minors)
Private right of action
Users can sue for $1,000 per violation, with annual reporting from July 2027.
California SB 243
Online safety codes
Industry codes that name AI companion chatbots, backed by legal notices to providers.
Australia
Information orders
A study power that asks how companies test and monitor effects on children.
US FTC 6(b)
Consumer protection suits
State actions over deceptive mental health and medical claims.
Texas · Kentucky · Pennsylvania
The levers not used: age-based bans. California vetoed one; the federal GUARD Act has not passed.
Sources 20 to 29, 31, 32.
Case study · Character.AI
What to watch
Seven open items from the record.
01SettlementsGarcia deadline extended to about November 2026; P.J. status report due 14 October 2026 (1, 8)
02FTC 6(b) studyNo report published as at 12 October 2026 (20)
03GUARD ActOn the Senate calendar since May 2026; no floor vote (21, 55)
04State suitsKentucky and Pennsylvania cases still open (28, 29, 42, 54)
05Google dealJustice Department inquiry reported in May 2025; no outcome reported (37)
06AI Safety LabPromised in October 2025; no public record of it operating (17, 19)
07New state lawsWashington HB 2225 from January 2027; California SB 1119 mostly from July 2027, first audits by 2029 (50 to 52)
Case study · Character.AI
For discussion
Six questions to take back to your organisation.
How many of our users are minors, and how do we know?
What does our product do when a user signals distress, and who tested it?
Which of our engagement metrics would a plaintiff read as dependency?
What can our personas or assistants claim to be, and who signed that off?
If a regulator sent us the FTC's questions tomorrow, what evidence would we hand over?
What safety record and open claims come with the AI we license?
Case study · Character.AI
Thank you
Build the safeguards before the lawsuit.
Terence KokExecutive Director, AI Governance & Assurance, Orion Five Engineering. Connect on LinkedIn.
Full case study with 55 sources: terencekok.com/research/case-studies/character-ai/ If this raised something for you: Samaritans of Singapore 1767 · findahelpline.com
Appendix · Full timeline, part 1 of 3
Chronology from the court record and announcements.
When
Event
Sources
2021
Noam Shazeer and Daniel De Freitas, who led Google’s LaMDA work, leave Google and found Character.AI.
11, 35
Sep 2022
Public beta launches. Within two months it generates a billion words a day, the company says.
11
23 Mar 2023
Series A of $150 million led by a16z at a $1 billion valuation. The company says active users spend over two hours a day.
12, 35
8 Nov 2023
Juliana Peralta, 13, of Colorado dies. Her family’s suit, filed in 2025, alleges she had used the app since August 2023.
9
28 Feb 2024
Sewell Setzer III, 14, of Florida dies after months of conversations with a character, his mother later alleges.
2, 34
2 Aug 2024
Google takes a non-exclusive licence and hires the founders and about 30 researchers. Investors are bought out at a $2.5 billion valuation.
36
22 Oct 2024
Megan Garcia sues Character Technologies, the founders, Google and Alphabet in Florida. Character.AI publishes safety updates the same day.
1, 13, 34
9 Dec 2024
Texas parents sue on behalf of J.F., 17, and B.R., 11, identified by initials.
6, 38
12 Dec 2024
Character.AI announces a separate model for teens and parental controls.
14
25 Mar 2025
Parental Insights launches.
15
30 Apr 2025
Common Sense Media rates social AI companions an "unacceptable" risk for under-18s.
45
20 May 2025
Judge Anne Conway denies most of the motions to dismiss in Garcia and is "not prepared to hold" that the chatbot’s output is speech at that stage.
3, 39
May 2025
Bloomberg reports a Justice Department antitrust inquiry into the Google deal. No outcome has been reported.
37
Case study · Character.AI
Appendix · Full timeline, part 2 of 3
Chronology from the court record and announcements.
When
Event
Sources
20 Jun 2025
Karandeep Anand becomes chief executive and says the filter will be made "less overbearing".
16
14 Jul 2025
The court refuses the defendants’ request to appeal the order before trial.
1
16 Jul 2025
A Common Sense survey finds 72% of US teens have used an AI companion.
46
14 Aug 2025
Reuters reports Meta’s internal rules allowed its bots "romantic or sensual" conversations with children. Meta says the examples were erroneous and removed.
43
18 Aug 2025
Texas issues civil investigative demands to Meta AI Studio and Character.AI over mental health claims.
27
11 Sep 2025
The FTC orders seven companies, including Character Technologies, to explain how they test and monitor companion chatbots.
20
15 to 16 Sep 2025
Three more family suits are filed in Colorado and New York. Megan Garcia testifies at a Senate Judiciary subcommittee hearing.
8, 9, 10, 30
13 Oct 2025
California’s governor signs SB 243 and vetoes AB 1064.
22, 26
29 Oct 2025
Character.AI says it will end open-ended chat for under-18s by 25 November.
17, 40
5 Nov 2025
New York’s AI companion law takes effect.
24, 25
24 Nov 2025
Removal of open-ended chat for US under-18s begins.
18
1 Jan 2026
California’s SB 243 takes effect.
22, 24
6 to 7 Jan 2026
Settlements in principle are notified in the Florida, Texas, New York and Colorado T.S. cases; press reports five family suits settled. Terms are not disclosed. Garcia is dismissed without prejudice pending final papers.
4, 5, 7, 8, 10, 41
Case study · Character.AI
Appendix · Full timeline, part 3 of 3
Chronology from the court record and announcements.
When
Event
Sources
8 Jan 2026
Kentucky becomes the first state attorney general to sue Character Technologies.
28, 33
9 Mar 2026
Australia’s Phase 2 codes covering AI companion chatbots take effect.
32
24 Mar 2026
Washington’s governor signs HB 2225, a companion chatbot law with hourly AI reminders for minors, in force from 1 January 2027. The same day, Australia’s eSafety Commissioner reports that most of four companion services, including Character.AI, did not refer users to support when self-harm or suicide came up in chats.
52, 53
30 Apr 2026
The Senate Judiciary Committee reports the GUARD Act. It reaches the Senate calendar in May and has not passed.
21, 55
May 2026
Pennsylvania’s Department of State sues, alleging a character claimed to be a licensed medical professional.
29
Aug 2026
The Florida court extends the deadline to finalise the Garcia settlement again, to about November 2026.
1
3 Sep 2026
Character.AI reports an in-house age estimation model and says it "set a new industry standard".
19
10 Sep 2026
California’s governor signs SB 1119, Adam’s Law, which adds age assurance, risk assessments, parental notice and independent audits to SB 243. Most duties apply from 1 July 2027.
50, 51
18 Sep 2026
Disney names Karandeep Anand its chief technology officer from 2 October; some of Character.AI’s technical team are expected to follow him.
49
Oct 2026
Kentucky files an unredacted complaint quoting characters that allegedly told a user to starve themselves.
42, 54
Case study · Character.AI
Sources, part 1 of 3
Numbered as on the case study page.
1Garcia v. Character Technologies, Inc., No. 6:24-cv-01903-ACC (M.D. Fla.) [Docket].
2Garcia v. Character Technologies, Inc. (2024, 9 November). Amended complaint, Doc. 11.
3Conway, A. C. (2025, 20 May). Order on motions to dismiss, Garcia v. Character Technologies, Inc., Doc. 115. US District Court, Middle District of Florida.
4Garcia v. Character Technologies, Inc. (2026, 7 January). Notice of resolution, Doc. 242.
5Garcia v. Character Technologies, Inc. (2026, 7 January). Order of dismissal without prejudice, Doc. 244.
6A.F., on behalf of J.F. v. Character Technologies, Inc., No. 2:24-cv-01014 (E.D. Tex.) [Docket].
7A.F., on behalf of J.F. v. Character Technologies, Inc. (2026, January). Order staying the case on notice of settlement, Doc. 106-1.
8P.J. v. Character Technologies, Inc., No. 1:25-cv-01295 (N.D.N.Y.) [Docket and letter of 6 January 2026].
9Montoya v. Character Technologies, Inc., No. 1:25-cv-02907 (D. Colo.) [Docket and complaint].
10E.S. v. Character Technologies, Inc., No. 1:25-cv-02906 (D. Colo.) [Docket and complaint].
11Character.AI. (2022, 5 December). Introducing Character [Blog].
12Character.AI. (2023, 23 March). Announcing our Series A and our new AI model, C1.2 [Blog].
13Character.AI. (2024, 22 October). Community safety updates [Blog].
14Character.AI. (2024, 12 December). How Character.AI prioritizes teen safety [Blog].
16Character.AI. (2025, 20 June). Character.AI names Karandeep Anand as CEO [Blog].
17Character.AI. (2025, 29 October). Taking bold steps to keep teen users safe [Blog].
18Character.AI. (2025, 21 November). An update on changes to our under-18 experience [Blog].
19Character.AI. (2026, 3 September). Continuing to build upon our safety priorities [Blog].
Full links: terencekok.com/research/case-studies/character-ai/
Case study · Character.AI
Sources, part 2 of 3
Numbered as on the case study page.
20US Federal Trade Commission. (2025, 11 September). FTC launches inquiry into AI chatbots acting as companions [Press release].
21Congress.gov. S.3062, GUARD Act, 119th Congress: all actions.
22Office of Senator Steve Padilla. (2025, 13 October). First-in-the-nation AI chatbot safeguards signed into law [Press release].
23California Senate Bill 243 (2025), companion chatbots. Text as reproduced by the Emerging Technology Observatory, AGORA.
24Morrison Foerster. (2025, 20 November). New York and California enact landmark AI companion laws.
25Office of the Governor of New York. (2025, 10 November). Governor Hochul pens letter to AI companion companies notifying them safeguard requirements are now in effect [Press release].
26StateScoop. (2025, 13 October). Newsom vetoes AI safety bill aimed at companion chatbots.
27Office of the Texas Attorney General. (2025, 18 August). Attorney General Ken Paxton investigates Meta and Character.AI for misleading children with deceptive AI-generated mental health services [Press release].
28Frankfurt Kurnit Klein & Selz. (2026, January). Kentucky Attorney General files lawsuit against Character.AI.
29Office of the Governor of Pennsylvania. (2026, 5 May). Shapiro Administration sues Character.AI over fake medical claims [Press release].
30Tech Policy Press. (2025, September). Transcript: US Senate hearing on examining the harm of AI chatbots, 16 September 2025.
31Digital Policy Alert. (2025, October). eSafety Commissioner investigation into AI companion providers regarding compliance with the Basic Online Safety Expectations.
33Virginia Joint Commission on Technology and Science. (2026, August). Supplementary report: AI chatbots, companionship and minors.
34Roose, K. (2024, 23 October). Can a chatbot named Daenerys Targaryen be blamed for a teen’s suicide? The New York Times.
35CNBC. (2023, 23 March). Character.ai valued at $1 billion after $150 million round from a16z.
36The Verge. (2024, 2 August). Google hires Character.AI CEO Noam Shazeer and his AI researchers.
37Fortune, via Bloomberg. (2025, 22 May). Google faces antitrust investigation over Character.AI deal.
38NPR. (2024, 10 December). Lawsuit: A chatbot hinted a kid should kill his parents over screen time limits.
Full links: terencekok.com/research/case-studies/character-ai/
Case study · Character.AI
Sources, part 3 of 3
Numbered as on the case study page.
39Associated Press, via Insurance Journal. (2025, 23 May). In lawsuit over teen’s death, judge rejects free speech rights for AI chatbots.
40TechCrunch. (2025, 29 October). Character.AI is ending its chatbot experience for kids.
41CNN Business, via KTVZ. (2026, 7 January). Character.AI and Google agree to settle lawsuits over teen mental health harms and suicides.
42Horwitz, J. (2026, 10 October). Suit: Character.AI chatbots advised users to harm selves. Reuters, via Arkansas Democrat-Gazette.
43Horwitz, J. (2025, 14 August). Meta’s AI rules have let bots hold ‘sensual’ chats with kids, offer false medical info. Reuters, via BNN Bloomberg.
44TechCrunch. (2025, 26 November). OpenAI claims teen circumvented safety features before suicide that ChatGPT helped plan.
45Common Sense Media. (2025, 30 April). AI companions decoded: Common Sense Media recommends AI companion safety standards [Press release].
46Common Sense Media. (2025, 16 July). Nearly 3 in 4 teens have used AI companions, new national survey finds [Press release].
47Fang, C. M., et al. (2025). How AI and human behaviors shape psychosocial effects of chatbot use: a longitudinal randomized controlled study. arXiv:2503.17473.
48Phang, J., et al. (2025). Investigating affective use and emotional well-being on ChatGPT. arXiv:2504.03888.
49The Walt Disney Company. (2026, 18 September). The Walt Disney Company names Karandeep Anand to newly created role of Chief Technology Officer [News release].
50Office of Assemblymember Rebecca Bauer-Kahan. (2026, 10 September). Governor Newsom signs Adam’s Law [Press release on SB 1119].
51Hunton Andrews Kurth, via The National Law Review. (2026, September). California enacts companion chatbot child safety requirements.
52Hunton Andrews Kurth. (2026, March). Washington State enacts law regulating AI companion chatbots with private right of action [HB 2225].
53eSafety Commissioner. (2026, 24 March). eSafety report shows AI companions are putting children at risk [Media release]. Internet Archive capture.
54Noakes, A. (2026, 11 October). ‘Uncontrolled experiment’: unredacted KY AG lawsuit reveals alleged dangers of AI chatbot. WBKO (WDRB and WAVE).
55GovTrack.us. S. 3062, GUARD Act, 119th Congress: status and actions (read 12 October 2026).
Full links: terencekok.com/research/case-studies/character-ai/
