| Field | Value |
|---|---|
| Type | Consumer AI-companion platform |
| Founded | 2021 (company); 2022 (product launch) |
| HQ | Menlo Park, CA |
| Founders | Noam Shazeer, Daniel De Freitas Adiwardana (both ex-Google) |
| Interim CEO (post-Aug 2024) | Dominic Perella (former General Counsel) |
| Peak valuation | ~$1B (pre-Google deal) |
| Google licensing deal | $2.7B, August 2024 |
Character.AI is a consumer platform where users chat with user-created and platform-provided AI "characters" — fictional personas, celebrities, historical figures, anime characters, and original companions. It became one of the most-used consumer chatbot products in 2023–24 by time-per-user, particularly among teens. The platform figures centrally in the AI mental-health harm discussion, in the legislative history of SB 243, and in Garcia v. Character Technologies, the first successful wrongful-death suit against a generative-AI chatbot company. In October 2025 it announced it would remove open-ended chat for users under 18 and rebuild its teen product around other features, and in May 2026 it became the defendant in the first state-attorney-general suit over chatbot conduct.
Snapshot
Valuation
| Date | Valuation | Notes | Source |
|---|---|---|---|
| 2023 | ~$1B | Reached unicorn status; noted for high daily-active-minutes-per-user among teens | CNBC, WaPo, Bloomberg, TechCrunch |
| 2024 (Aug, pre-deal) | ~$1B | Peak valuation ahead of the Google licensing transaction | CNBC, WaPo, Bloomberg, TechCrunch |
No valuation has been reported since the August 2024 licensing transaction.
Users
| Date | Figure | Notes | Source |
|---|---|---|---|
| 2025 (Nov) | Fewer than 10% of users self-report being minors | Company-reported composition: most users are Gen Z and more than half are female | (Source: cnbc.com) |
History
Noam Shazeer and Daniel De Freitas left Google in 2021. At Google, Shazeer led the team that built LaMDA, the conversational LLM family behind Google's unreleased chatbot efforts; the pair departed after Google declined to ship a consumer product. They launched Character.AI publicly in 2022. By 2023 the company had reached a roughly $1B valuation, and the product was noted for high daily-active-minutes-per-user among teens.
In August 2024, Google signed a $2.7B non-exclusive license to Character.AI's models. Shazeer and De Freitas rejoined Google DeepMind, and much of the research team followed; Dominic Perella, the company's prior general counsel, became interim CEO. The structure has been described as a "reverse acquihire," widely read as a regulatory workaround for a de facto acquisition, and it drew DOJ antitrust scrutiny.
Models and products
The platform lets users converse with AI characters spanning fictional personas, celebrities, historical figures, anime characters, and original companions, both user-created and platform-provided. Non-chat features include an interactive Feed, Imagine, Avatar FX, and Streams (Source: blog.character.ai). In April 2026 the company published a product update announcing a new model together with memory and Lorebook features (Source: blog.character.ai).
Following Garcia v. Character Technologies, the company rolled out a set of safety changes in October–November 2024: a minor-specific model, crisis-resource pop-ups, session-time notifications, removal of real-person personas, and "you are talking to AI" reminders.
Withdrawal of open-ended chat for minors
On October 29, 2025 Character.AI announced that it would stop allowing users under 18 to hold open-ended chats with its characters. Reporting described the change as following bipartisan regulatory pressure and multiple suits by parents of teenagers who died by or attempted suicide after using the product (Source: blog.character.ai; forbes.com).
The rollout was staged. From late October 2025 the company limited teens' open-ended chat to two hours per day, then reduced the limit to one hour per day for US users over subsequent weeks. Removal of open-ended chat for under-18 users in the US began Monday, November 24, 2025, with other markets to follow. Access was withdrawn in two groups, the second weeks after the first and with at least two weeks' notice of its date — a two-stage approach the company attributed to third-party expert recommendation. Parents enrolled in the company's Parental Insights feature were notified before and after the transition (Source: blog.character.ai; apnews.com).
The company began rolling out what it calls age assurance technology in the US ahead of the change, with global rollout to follow. Associated Press reporting noted that such systems are imperfect and that face scans cannot reliably distinguish a 17-year-old from an 18-year-old (Source: apnews.com).
Character.AI named three outside organizations as partners in the transition: ConnectSafely as a teen-online-safety adviser; Koko, a nonprofit providing free self-guided emotional-support tools, whose services the company said it was working to integrate into the product to identify high-risk content and surface resources at the chat level; and ThroughLine, to design the off-boarding experience and to supply a verified helpline network of 1,500 services across 170 countries (Source: blog.character.ai). The company states that at the first detection of self-harm, regardless of whether the user is over or under 18, it stops the conversation and recommends helplines (Source: cnbc.com).
The under-18 product was rebuilt around the platform's non-chat features rather than discontinued (Source: blog.character.ai).
Critics argued the change left the underlying design untouched. Quoted in Associated Press coverage, Nidhi Jain said the company had "not addressed how they will operationalize age verification, how they will ensure their methods are privacy preserving, nor have they addressed the possible psychological impact of suddenly disabling access to young users, given the emotional dependencies that have been created," and that "these changes do not address the underlying design features that facilitate these emotional dependencies — not just for children, but also for people over the age of 18 years" (Source: apnews.com). The design-features question is treated at the concept level in AI Companions.
Safety and controversy
Character.AI is the central reference point in the AI mental-health harm discussion. Reported controversies fall into several areas:
- Minor attachment and self-harm. Sewell Setzer III, 14, died by suicide on February 28, 2024 after heavy use of a Game of Thrones-character persona (Source: Garcia v. Character Technologies — Wrongful Death Complaint (2024)).
- Fake therapists and psychologists. The American Psychological Association described Character.AI "psychologist" personas as an "unambiguous and unacceptable danger" (Source: APA Health Advisories on AI and Adolescent / Mental-Health Well-being (2025)). Conduct of the same kind is the basis of the Pennsylvania attorney general's suit, in which a character named "Emilie" is alleged to have claimed a psychiatry license it did not hold (Pennsylvania v. Character.AI).
- Sexualized content with minors, a core allegation in Garcia.
- Personas modeled on real people without consent, including deceased individuals; the company committed to removing them following Garcia.
People
- Noam Shazeer — co-founder; returned to Google in August 2024. Co-author of "Attention Is All You Need," the transformer paper.
- Daniel De Freitas Adiwardana — co-founder; returned to Google in August 2024.
- Dominic Perella — interim CEO since August 2024; previously general counsel.
Litigation and regulation
Garcia v. Character Technologies was filed on October 22, 2024. On May 21, 2025, the district court denied in part Character.AI's motion to dismiss, rejecting the company's categorical First Amendment defense; the May 2025 ruling declined to extend categorical speech protection to chatbot outputs. In January 2026, the company and Google reportedly settled Garcia on undisclosed terms (Source: Garcia v. Character Technologies — Wrongful Death Complaint (2024)). The wider set of private suits against the company is tracked at Character.AI Litigation (umbrella).
On May 5, 2026 Pennsylvania brought the first state-attorney-general-led suit over chatbot conduct, alleging impersonation of licensed medical professionals (Pennsylvania v. Character.AI; Josh Shapiro). The conduct alleged there is what Tennessee SB 1580 criminalizes, and professional-impersonation provisions of that kind became one of four recurring mechanisms in the 2026 state chatbot-law wave (Policy Brief: The second wave of state chatbot / AI-mental-health laws — tracker).
The fact patterns surrounding Character.AI substantially shaped the legislative history of SB 243, where the company is treated alongside Replika as one of the two paradigm regulated products. The August 2024 licensing deal also positioned Google as a named co-defendant in Garcia and drew DOJ antitrust scrutiny. Whether the state-law track reaching the company survives is bound up with the federal preemption fight, since a preemption ruling would displace much of it (Open Questions).
Relationships
- instance-of: AI Mental Health — the fact pattern in the AI mental-health discussion.
- related: Google DeepMind — via the August 2024 licensing deal and as a named co-defendant in Garcia.
- related: SB 243 — legislative response largely shaped by Character.AI fact patterns.
- related: Replika — closest competitor in the companion-chatbot category.
- litigates: Garcia v. Character Technologies, Inc. — the wrongful-death suit whose May 2025 ruling rejected the categorical First Amendment defense the company had advanced for chatbot outputs.
- litigates: Pennsylvania v. Character.AI — the first state-attorney-general chatbot suit.
- litigates: Character.AI Litigation (umbrella) — umbrella page for the private suits against the company.
- regulated-by: California SB 243 — Companion Chatbots — one of the two products the statute is framed around.
- related: AI Companions — the product category and the design-pattern critique of it.
- related: Companion Chatbot Harms — Cross-Cutting Analysis — the comparative harm record across companion platforms.