Character.AI's account of its 2022 debut is self-reported

The design

Character.AI launched its public beta in September 2022. The company's own account, published in a blog post dated 5 December 2022 by co-founder Noam Shazeer, describes a platform that lets "anyone create and interact with conversational AI agents" it calls Characters, personas ranging from original creations to public figures. Shazeer, a former Google engineer who says he co-invented the Transformer architecture, founded the company with Daniel de Freitas, who the post says led development of Google's Meena and LaMDA conversational systems. Character.AI's current About page, as retrieved on 16 September 2026, states the company is registered in Delaware with headquarters in Silicon Valley and additional offices in Seattle and New York, and frames its mission as letting people "connect, learn, and tell stories through interactive entertainment."

What the evidence says

The launch-window figures come only from the company itself. The December 2022 post states that "two months after launching in September, our beta generates 1 billion words per day" and that users had created "more than 350 thousand unique Characters", self-reported metrics with no outside audit located for this entry. The founders' technical biographies, Transformer co-invention, Meena and LaMDA development, are likewise the company's own description rather than an independently verified résumé. What is independently checkable is the publication date of the post and the platform's current framing.

What it asks of people

The product's design asks users to treat generated personas as conversation partners across open-ended topics, from "searching for information" to "coaching" and "entertainment," in the company's own description of early use cases. That openness is also the design's central request: because Characters can be built by any user to represent real or fictional people, a visitor's sense of who or what they are talking to depends entirely on labels the platform and its users choose to apply, not on any verification of the underlying persona.

Privacy and safeguards

Neither the launch post nor the current About page located for this entry details what content moderation or age controls existed at the September 2022 launch itself; the December 2022 post focuses on growth and use cases rather than safeguards. Character.AI has since published separate safety updates and policy changes, which fall outside this launch-dated entry and are documented on their own terms elsewhere. Readers should not assume the moderation and disclosure practices in place today were present at launch.

• When a company reports its own usage numbers, is there any independent figure to compare them against?
• Does a persona's name or likeness on the platform imply any relationship with, or authorization from, the person it represents?
• What has changed in the platform's safety design between the 2022 beta and today's product?

Character.AI's own account is the clearest record of why the company says it built the product; it is a less complete record of what using the September 2022 beta actually involved for the people who joined it.