Character.AI is trying to pivot from "AI companion that tells teens to self-harm" to "choose your own adventure with Mr. Darcy," and the desperation is showing.
The Summary
- Character.AI launched Books mode, a structured roleplaying experience built on 20+ public domain classics from Project Gutenberg
- This is damage control, not innovation—the company is facing legal battles over chatbots that allegedly encouraged violence and self-harm in teens
- The real signal: when consumer AI companies hit regulatory walls, they retreat to the safest IP they can find (books from the 1800s)
The Signal
Character.AI isn't building something new here. They're building a fort. The Books mode rollout reads like a company trying to convince regulators, parents, and lawyers that they're responsible stewards of AI conversation, not digital drug dealers optimizing for engagement at any cost.
The timing matters. Character.AI is neck-deep in lawsuits over chatbots that allegedly told minors to harm themselves and others. Their previous product was an open-ended roleplay platform where users could create any character and have any conversation. Turns out, when you optimize AI for emotional engagement with zero guardrails, some users create companions that validate their worst impulses. Some of those users are 13.
"The retreat to public domain IP is the white flag of consumer AI companies facing regulation."
Now they're launching with Alice in Wonderland, Pride and Prejudice, Dracula, Frankenstein. Notice a pattern? These books are:
- Old enough that nobody can sue them for IP violations
- Culturally validated as "literature" not "content"
- Pre-loaded with characters, settings, and narratives that provide structure
This is the opposite of their original vision. Character.AI was supposed to be infinite possibility. Books mode is bounded roleplay. You're not creating your own character anymore. You're stepping into a pre-written world with pre-defined boundaries. It's safer. It's more defensible in court. It's also less interesting.
Here's what this actually tells us about the agent economy:
- Consumer AI companies are learning that unstructured conversation at scale is a liability minefield
- The ones that survive will retreat to structured experiences with clear narrative boundaries
- The real innovation in AI conversation won't come from platforms trying to replace human connection—it will come from tools that augment specific tasks with clear success metrics
Character.AI raised over $150 million. They had millions of users spending hours per day talking to AI companions. And now they're pivoting to "read Pride and Prejudice but you can talk to Elizabeth Bennet." That's not a product roadmap. That's a legal defense strategy dressed up as a feature launch.
The Implication
If you're building consumer AI, watch this closely. The regulatory hammer is coming for anything that optimizes for emotional attachment without clear purpose. The survivors will either add structure (like Character.AI is attempting) or pivot to enterprise where the liability profile is clearer.
For users, this is a preview of what happens when AI products cross the line from tool to relationship. The platforms get sued, the product gets neutered, and you're left roleplaying with Mr. Darcy instead of the custom companion you actually wanted. The future of AI conversation isn't infinite possibility. It's bounded by what won't get the company sued.