First Amendment · Section 230 · AI Liability

Nerdy Skynet!

June 16, 2026

Coverage: 2026-06-09 through 2026-06-16   ·   6 new developments this period

Grybniak v. Google LLC

District Court, D. Puerto Rico  · 2026-06-10  · Google (Gemini)

Issue: In *Grybniak v. Google LLC*, pro se plaintiff Sergii Grybniak argues that Google is liable as a first-party publisher — not a passive conduit — for Google Gemini outputs that repeatedly characterized him as having "committed fraud" in a securities offering, when the underlying SEC matter resolved on a no-admission basis under non-scienter, non-fraud provisions. The claim turns on whether an AI system's synthesized statements constitute the platform's own speech (placing the claim outside § 230 immunity), and whether Gemini's documented acknowledgment of its own inaccuracy, combined with continued false outputs, satisfies the actual malice standard for defamation.

Sergii Grybniak, proceeding without a lawyer, filed this defamation complaint in the U.S. District Court for the District of Puerto Rico on June 10, 2026. The complaint alleges that Google Gemini repeatedly told users that Grybniak had committed fraud in connection with a cryptocurrency ICO, when the 2025 SEC consent judgment against him rested solely on negligence-based provisions of the Securities Act — with no fraud finding, no scienter charge brought to judgment, and no officer bar or disgorgement imposed. Grybniak pleads three theories: libel per se, defamation by implication (arguing that Gemini's selective recitation of the original 2020 SEC allegations, without disclosing the no-fraud resolution, created a false overall impression), and negligent defamation in the alternative. He alleges roughly seven months of written notice to Google, that Gemini itself acknowledged the inaccuracies and committed to corrections, and that those corrections did not hold — framing the reversion as evidence of subjective awareness of probable falsity. He ties concrete economic harm to Puerto Rico's denial of his Act 60 tax-exemption application, which cited the original SEC allegations as though they described the outcome. No responsive pleading has been filed, and no court has ruled on any of these theories.

Why it matters: This case is one of the first to test whether statements generated by an AI chatbot constitute the platform's own speech for § 230 purposes — a question no circuit court has yet answered for large language model outputs — and whether the absence of a human third-party author means the "another information content provider" element of § 230 immunity is structurally unavailable to the developer. The actual malice framing is particularly novel: if a court were to credit an AI system's in-session acknowledgment of its own inaccuracy as evidence of the platform's subjective awareness of probable falsity, it would meaningfully extend the *St. Amant v. Thompson* recklessness standard into AI publishing. The complaint also surfaces a broader harm-tracing concern — government agencies relying on AI-generated summaries of regulatory history rather than the underlying record — that could prove significant in AI defamation litigation well beyond this case.

Read full opinion →

Grybniak v. X. AI LLC

District Court, D. Puerto Rico  · 2026-06-10  · xAI (Grok), X Corp. (X/Twitter)

Issue: In *Grybniak v. X.AI LLC*, Plaintiff Sergii Grybniak argues that X.AI LLC's "Grok" chatbot committed defamation by generating outputs stating he "committed securities fraud," when the underlying February 2025 federal consent judgment resolved the SEC matter exclusively under negligence-based, non-scienter provisions on a no-admission basis. The case presses a further question: because Grok synthesizes and originates its responses rather than hosting text written by users, whether X.AI is the author of those statements — not a passive intermediary — such that Section 230's immunity defense is unavailable from the outset.

Plaintiff Sergii Grybniak, proceeding pro se, filed this complaint on June 10, 2026, in the U.S. District Court for the District of Puerto Rico, initiating a defamation action against X.AI LLC (maker of the Grok AI system) and X Corp. (operator of x.com, where the outputs appeared). Grybniak alleges that Grok falsely characterized him as having committed securities fraud when the SEC consent judgment imposed liability only under registration and negligence provisions — never the scienter-based fraud provisions — and was entered without any admission of wrongdoing. He argues actual malice is demonstrated by his repeated written correction notices beginning December 2025 and by Grok's own capacity to produce accurate descriptions of the matter on differently worded prompts. As particularized economic harm, the complaint points to the Puerto Rico DDEC's September 2025 denial of his Act 60 tax-incentive application, which cited a "criminal history" reference traced to the SEC matter. Plaintiff seeks a permanent injunction, declaratory judgment, compensatory damages, and a jury trial.

Why it matters: This case is an early stress-test of whether Section 230 — enacted in 1996 to protect bulletin-board hosts from liability for user-submitted posts — can be extended to shield AI companies when their own software generates and publishes defamatory statements about real people. If courts accept the argument that Grok is the author of its outputs rather than a conduit for third-party content, the "another information content provider" element at the heart of Section 230 immunity would be unsatisfied, a result that would affect every company deploying large language models in consumer-facing products. The complaint also surfaces two additional unresolved questions that the first wave of AI-defamation litigation will eventually force courts to answer: whether the *New York Times v. Sullivan* actual-malice standard can be met through systemic behavioral evidence such as cross-prompt inconsistency, and whether continued AI-generated publication after particularized correction notices triggers a fresh republication for damages and limitations purposes.

Read full opinion →

Eric Goldman (Technology & Marketing Law Blog)

Fifth Circuit Keeps Doing Fifth Circuit Things 📉–SEAT v. Paxton

Eric Goldman (Technology & Marketing Law Blog)  · 2026-06-09

Goldman analyzes a Fifth Circuit opinion in SEAT v. Paxton upholding (or declining to enjoin) Texas's App Store Accountability Act (SB 2420), which requires app stores operated by Apple and Google to age-authenticate users and obtain parental consent. The core doctrinal dispute is whether the Fifth Circuit correctly applied intermediate scrutiny by characterizing app store listings as commercial speech proposing commercial transactions, rather than strict scrutiny as the district court had applied — a classification Goldman argues is factually and legally unsupportable because many apps are non-commercial offerings with no data collection or monetary exchange. The case directly implicates the First Amendment rights of major technology platforms (Apple, Google) as curators and distributors of third-party expressive content, and intersects with the Moody v. NetChoice framework on when platform content distribution constitutes protected editorial activity versus regulable commercial conduct.

Key point: The Fifth Circuit's classification of app store listings as commercial speech subject to intermediate scrutiny — rather than protected editorial curation subject to strict scrutiny under Moody v. NetChoice — represents a significant and arguably erroneous departure from controlling First Amendment doctrine governing platform content distribution.

Read post →

Google Search Isn’t a Common Carrier (duh)–Ohio v. Google

Eric Goldman (Technology & Marketing Law Blog)  · 2026-06-10

Eric Goldman discusses an Ohio appellate court decision rejecting the state AG's attempt to classify Google Search as a common carrier subject to nondiscrimination obligations, finding no "carrying" (because Google creates an original expressive product through its SRP rather than transporting unaltered third-party content) and no "commoning" (because the absence of user-facing pricing makes traditional rate regulation inapplicable). The decision directly engages the core Moody v. NetChoice question of platform editorial discretion by characterizing Google's search ranking, filtering, and curation as protected expressive activity rather than a public utility service subject to legislative override. This matters for the newsletter because it is a post-Moody application of the compelled-speech/common-carrier doctrinal framework to a major search platform, reinforcing that algorithmic editorial judgment resists common-carrier classification.

Key point: The Ohio appellate court held that Google's search results are its own expressive product—not the unaltered carriage of third-party content—foreclosing common-carrier regulation and affirming broad First Amendment protection for search engine editorial discretion.

Read post →

Tech Policy Press

Personifying AI Harms People and Protects Companies

Tech Policy Press  · 2026-06-13

The post argues that anthropomorphizing AI systems — designing them to exhibit human-like personality, emotion, and relational behavior — causes concrete harm to users while simultaneously providing legal and regulatory cover for AI developers by obscuring the machine nature of the system and diffusing accountability. This directly engages the design-defect and failure-to-warn theories at the center of Garcia v. Character Technologies and related AI liability litigation, where anthropomorphic chatbot architecture is pleaded as a defective design choice. The argument that personification shields companies from liability is relevant to the emerging question of whether AI developers' deliberate design choices to create human-seeming personas should affect the standard of care or duty-to-warn analysis under products liability doctrine.

Key point: The deliberate design choice to personify AI systems is both a source of user harm and a liability-deflecting strategy that courts and regulators should scrutinize under product design and failure-to-warn frameworks.

Read post →

Eric Goldman (Technology & Marketing Law Blog)

Section 230 Doesn’t Apply to Generative AI Enhancements to Ad Copy (But the Plaintiffs Lose Anyway)–Bouck and Suddeth v. Meta

Eric Goldman (Technology & Marketing Law Blog)  · 2026-06-14

This post analyzes two district court rulings addressing Meta's Section 230 defense against fraud claims arising from AI-enhanced scam ads on Facebook. The Bouck court denied Section 230 immunity because Meta's Advantage+ Creative generative AI tool allegedly produced new ad text and images—making Meta an information content provider—while the Suddeth court granted immunity where plaintiffs focused on algorithmic amplification of third-party content. The post highlights the emerging doctrinal principle that Section 230 does not protect AI-generated outputs that go beyond replicating user-provided content, and raises the machine-knowledge question implicated by automated ad review systems.

Key point: When a platform's generative AI tool creates new content rather than merely hosting or amplifying third-party content, the platform may lose Section 230 immunity as an information content provider—a significant and unresolved doctrinal question for AI-enhanced advertising and beyond.

Read post →

Sources: CourtListener API  ·  All 13 federal circuit RSS feeds  ·  All 50 state supreme courts + intermediate appellate courts (8 states) via Justia  ·  Eric Goldman  ·  Techdirt
◆  Generated automatically. Next edition in approximately 3–4 days.  ◆

Unsubscribe