Week 2.5 - Open Letter to Jason Kuo
March 12, 2026
Hello Jason!
For some reason, you find my project “interesting” (a word you use with the same loose vagueness the words “truth” or “justice” carry). You asked for the stuff I’m reading. I thought it fitting to share my findings with my other readers, so I hope you don’t mind my communication in this form. Here’s a small compilation of most the information that I’ve found.
First, I recommend reading the two blog posts I’ve made on the topic:
The case Garcia v. Character Technologies has a lot of interesting court documents that you can sift through. Finding what’s thought provoking and what’s boring legal procedure can be a tad time consuming; thankfully I’ve already done that for you.
This site has all the public court filings: https://www.courtlistener.com/docket/69300919/garcia-v-character-technologies-inc/
I’ll be referring to court filings based on their document number for easy reference. The document numbers correspond to the number found on the left hand side of the docket entries on the site’s home page. (You can also just click on the links I’ve made for you here).
Here’s the original complaint filed by the plaintiff (Document #1). There’s some very powerful language in accusing the defendants here. I loved reading every word of it (although the judge thinks it’s disorganized…)
You can look at the attached exhibits by clicking on the highlighted button below.
See also the plaintiff’s second amended complaint (Document #11) and the third amended complaint (Document #149).
The complaint is long. Most of it explains the plaintiff’s understanding of the C.AI product. Skim Section V.B, ¶ 79-97 for evidence of the alleged negligence of the individual defendants Shazeer and De Freitas. If you read my first blog post, then you can skip Section V.D. Skim all of Section V.E and Section V.F ¶ 279-283 for evidence of the alleged negligence of the company.
Each of the defendants had their own reasons for believing why they don’t deserve to be prosecuted. For example, the owners of Character.AI believe that the limited liability (the LL in LLC) protects them from liability; in addition, they believe the court lacks jurisdictional authority to hear their case. Google alleges they simply provided Character.AI with cloud services to produce the product, falling short of evidence to implicate them in aiding and abetting. The arguments are more complex than this, obviously, so if you’re interested in that too (more corporate law than First Amendment), then you can look here:
DANIEL DE FREITAS’S MOTION TO DISMISS (Document #63)
NOAM SHAZEER’S MOTION TO DISMISS (Document #65)
CHARACTER TECHNOLOGIES, INC.’S MOTION TO DISMISS (Document #59)
GOOGLE LLC’S AND ALPHABET INC.’S MOTION TO DISMISS (Document #61)
You can also look at the plaintiff’s responses:
PLAINTIFF’S CONSOLIDATED RESPONSE IN OPPOSITION TO NOAM SHAZEER’S AND DANIEL DE FREITAS’ MOTIONS TO DISMISS (Document #84)
PLAINTIFF’S RESPONSE IN OPPOSITION TO CHARACTER TECHNOLOGIES, INC.’S MOTION TO DISMISS (Document #85)
PLAINTIFF’S RESPONSE IN OPPOSITION TO GOOGLE LLC’S MOTION TO DISMISS (Document #86)
And the defendants’ responses to that:
CHARACTER TECHNOLOGIES, INC.’S REPLY IN SUPPORT OF ITS MOTION TO DISMISS PLAINTIFF’S FIRST AMENDED COMPLAINT (Document #98)
GOOGLE LLC’S AND ALPHABET INC.’S REPLY IN SUPPORT OF ITS MOTION TO DISMISS (Document #99)
DANIEL DE FREITAS AND NOAM SHAZEER’S REPLY IN FURTHER SUPPORT OF THEIR RESPECTIVE MOTIONS TO DISMISS (Document #100)
Esteemed amici curiae also weighed in to give their opinions on the motions to dismiss, supporting the plaintiff:
I personally recommend reading all of Character.AI’s (related) filings and Daniel De Freitas initial filing to dismiss (his lawyer’s a bit sassy in the jurisdictional argument to dismiss). Google’s is quite tame in comparison, since they really aren’t in the main crosshairs of this case’s allegations.
What’s interesting to my project specifically is the defendants’ motion to dismiss on first amendment grounds. They all make similar arguments, but I thought it important to read all the documents to ensure I didn’t miss some groundbreaking argument. I’m still getting through it, so maybe you can send me a section you thought was interesting if you thought I may have missed it.
The judge ruled on their motions to dismiss here. The judge was not sympathetic to the defendants’ arguments, and quashed their motion:
Order on Motion to Dismiss (Document #115).
Following the judge’s motion, Character.AI filed for immediate appeal (i.e. interlocutory appeal) (Document #130) under 28 U.S.C. § 1292(b), which is rarely afforded. Google also joined Character.AI in their immediate appeal (Document #132). Motions to dismiss are not appealable under normal circumstances, which is why the company moved to appeal under this more rigorous avenue. Three prongs must be met for immediate appeal to be granted:
1. The order “involves a controlling question of law”
2. “there is “substantial ground for difference of opinion” as to that question and”
3. ““an immediate appeal from the order may materially advance the ultimate termination of the litigation.””
Here’s the plaintiff’s response:
Esteemed amici curiae also weighed in to give their opinions on the interlocutory appeal:
In support of defendants:
Amicus Curiae Brief in Support of Defendant Character Technologies Inc. by Eugene Volokh, Jane Brambauer (Document #133-1)
Amicus brief of CDT & EFF by Electronic Frontier Foundation (Document #145-1)
Brief of Amicus Curiae Foundation for Individual Rights and Expression in Support of Certification of Immediate Appeal (Document #148-1)
Brief of Amici Curiae NetChoice and Chamber of Progress in Support of Character Technologies Inc.’s Motion for Certification of Immediate Appeal (Document #151-1)
The Court denied their motion again here (Document #167). The Court denied Character.AI’s motion under the third prong, citing the plaintiff’s arguments attacking the “design features and functionalities” of the C.AI product. This would trigger intermediate scrutiny, as opposed to strict. The question of First Amendment in relation to the ideas and information is therefore not applicable, as what is being changed is now the way in which speech is disseminated, not the speech itself.
That’s the Court’s reasoning, at least. I’m not sure if I agree, since changing the way the LLM is trained will undoubtedly change the outputs it produces. Either way, the Court explains further that even if the appeal was granted, the case would still proceed. The parties would still need to litigate what scrutiny should apply, and subsequently to what extent First Amendment exceptions should apply. Because these questions would still need to be answered (through litigation), answering the question would not “materially advance the ultimate termination” of litigation.
So, the motion is quashed, and the questions are unanswered. The Court believes these questions are more fitting for the appellate review; though, we know it won’t reach that point since the casse is currently being settled.
Hope this satisfies your curiosity for a while…

Leave a Reply
You must be logged in to post a comment.