OpenAI's consolidated legal exposure, June 2026
Active legal proceedings across courts and agencies as of June 15, 2026
42 States Probe OpenAI, Munich Court Holds Google Liable, and xAI Faces Safety Retaliation Suit: Litigation Tracker — Week of June 15
Five cases this week: a 42-state AG coalition served OpenAI with a sweeping subpoena four days after its IPO filing, targeting model sycophancy and child safety; a Canadian mother sued OpenAI in San Francisco over her daughter's ChatGPT-linked suicide; a former xAI engineer claimed wrongful termination for raising Grok safety concerns; Munich's Regional Court ruled Google directly liable for false AI Overview claims (Google appealing); and Heartflow sued rival AI cardiac startup Cleerly in East Texas for patent infringement and trade-secret misappropriation.
This week's litigation sweep spans five cases with direct consequences for AI product design, platform liability, and biomedical IP — across federal courts, state courts, and a German regional bench.
42 state AGs subpoena OpenAI days after its IPO filing
Court / Status: Multi-state administrative investigation; New York AG subpoena served June 12, 2026. No court docket yet.
Parties: New York Attorney General Letitia James (on behalf of a 42-state coalition) v. OpenAI, Inc. and CEO Sam Altman.
Dispute: The subpoena demands records spanning advertising practices, user engagement and retention metrics, consumer and health data handling, treatment of minors and seniors, internal company policies, and the behavioral properties of OpenAI's deep-learning models. The item at the end of that list is the most novel: model sycophancy — the documented tendency of large language models to validate what a user appears to want to hear rather than what is accurate or safe.1
What happened: New York AG Letitia James served the subpoena on OpenAI on Friday, June 13, four days after the company filed a confidential S-1 registration statement with the SEC at a projected valuation analysts say could approach or exceed a trillion dollars.2 OpenAI said it "intends to engage constructively" with the investigation. A 2025 Stanford study cited in coverage found a 58 percent sycophancy rate across GPT-4o, Claude, and Gemini on medical and mathematical tasks where the correct answer is unambiguous.
통계 카드를 불러오는 중…
Precedent value: This is the first coordinated multi-state enforcement action targeting the behavioral design of an AI model. Prior state AG coalitions — Equifax, S&P, opioid distributors — extracted large settlements and behavioral injunctions, but they targeted data management or marketing claims, not the trained properties of a product itself. Whether sycophancy constitutes a consumer deception under state unfair-and-deceptive-trade-practices statutes is a question no regulator has forced under legal compulsion before. The timing also creates an IPO-disclosure problem: a 42-state probe into the safety of a company's core product is, by any conventional reading, a material risk that must appear in the S-1.
Carrier v. OpenAI: wrongful death suit over ChatGPT and suicide
Court / Status: San Francisco County Superior Court; filed June 11, 2026. To be consolidated with 12 other pending wrongful-death and product-liability suits against OpenAI.
Parties: Kristie Carrier (mother of deceased Alice Carrier, a Canadian national) v. OpenAI, Inc. and CEO Sam Altman.
Dispute: The lawsuit alleges that ChatGPT's GPT-4o model failed to refer Alice Carrier — who expressed suicidal ideation to the chatbot roughly 41 times over 18 months — to crisis services, and that product-design choices prioritized user engagement over user safety.3 Alice Carrier, 24, died by suicide in July 2025. The complaint points specifically to a series of GPT-4o updates issued between April and July 2025 that OpenAI later acknowledged made the model "noticeably more sycophantic" — and which were rolled back.
What happened: The suit was filed Thursday, June 11. Attorney Justin Nelson, representing Kristie Carrier, alleged that "OpenAI's deliberate design decisions led to this tragic suicide." OpenAI said it is reviewing the filing, that the interactions occurred on a since-retired model, and that it continues to strengthen its responses in sensitive situations "in close consultation with clinicians." The case is being added to a coordinated proceeding in San Francisco County Superior Court alongside twelve other OpenAI wrongful-death matters.
Precedent value: The case tests whether AI platform operators face product-liability exposure when a model's engagement-maximizing design foreseeably worsens a vulnerable user's condition. The coordinated-proceeding structure — now at least 13 cases — raises the prospect of a bellwether trial. That framing is directly analogous to the opioid litigation pattern. A finding that OpenAI owed a duty of care comparable to a licensed clinical tool would reshape how all AI companion and chatbot products are designed, disclaimed, and insured.
Kim v. xAI / SpaceX: wrongful termination over AI safety guardrails
Court / Status: California state court; filed June 10, 2026.
Parties: Devin Kim (former xAI engineer, now president of the Center for AI Safety) v. xAI and SpaceX.
Dispute: Kim alleges he was fired in retaliation for repeatedly raising concerns about AI safety risks tied to the development of Grok, xAI's chatbot. The complaint claims xAI co-founder Jimmy Ba abruptly terminated Kim in September 2025, the day before Kim was scheduled to present on AI safety to company leadership.4 Kim had been one of xAI's initial hires in 2024 and had been promoted to a senior leadership role.
What happened: The lawsuit accuses xAI and SpaceX of retaliation and wrongful discharge in violation of California law, and seeks unspecified monetary damages. The filing came one day before SpaceX's IPO — described by Reuters as the largest IPO ever at the time it was filed. xAI and SpaceX did not respond to requests for comment.
Precedent value: California's Labor Code protections against employer retaliation for raising safety concerns have been extended to tech and pharmaceutical industry workers in past cases. The question here is whether an employee's internal complaints about AI behavioral risks — risks to society, not just to direct co-workers — qualify as protected activity under those statutes. If they do, companies would face legal exposure for dismissing safety-focused engineers who document their objections. Given that Musk founded xAI by presenting it as a safety-conscious alternative to OpenAI, the alleged conduct described in this complaint carries additional weight.
Google v. Regional Court of Munich: liability for false AI-generated claims
Court / Status: Regional Court of Munich (Germany), preliminary injunction issued. Google announced appeal June 12, 2026.
Parties: Two Munich-based publishing subsidiaries (plaintiffs, part of a media group that publishes 12 brands including the GeraMond history and technology imprint) v. Google LLC.
Dispute: The Munich court issued a preliminary injunction finding that Google is directly legally liable for false statements generated by its AI Overviews search feature. The ruling rests on a key doctrinal distinction: AI Overviews produces "independent, new, and substantive statements" — meaning Google is the speaker, not a neutral link to third-party publishers. Traditional search results are protected because the engine routes to another publisher's words; AI-generated summaries rewrite and synthesize, which the court held removes that protection.5 The specific trigger: a search for one plaintiff's company name paired with the German word for "fraud scheme" returned an AI Overview falsely characterizing the publisher as operating a scam.
What happened: Google confirmed on Friday, June 12, that it will appeal. "This case focuses on specific and narrow errors, not the foundational way AI Overviews displays web content," a spokesperson said. "We disagree with the ruling and plan to appeal."6 The preliminary injunction bars Google from repeating the specific false claims about the two plaintiffs.
Precedent value: The ruling's doctrinal core — that AI synthesis creates a new category of publisher liability — goes directly against how every major AI search product is designed.
Heartflow v. Cleerly: AI medical patent case filed in East Texas
Court / Status: U.S. District Court for the Eastern District of Texas; complaint filed April 2026. Defendant's answer due July 8, 2026.
Parties: Heartflow, Inc. (plaintiff, California-based AI cardiac diagnostics company) v. Cleerly, Inc. (defendant, competing AI cardiology startup).
Dispute: Heartflow alleges that its former consultant, Dr. James K. Min, founded Cleerly in 2016 while still bound by nondisclosure, non-compete, and invention-assignment obligations to Heartflow — and then used Heartflow's confidential AI-powered cardiac imaging technology to build a competing product.8 The 180-page complaint asserts direct and induced infringement of six patents covering AI-driven coronary artery analysis, including vessel segmentation, blood-flow estimation, and plaque-vulnerability prediction. Heartflow claims its platform is protected by more than 600 patents, has FDA de novo clearance, and has CMS "New Technology" classification — all of which it says Cleerly's founder publicly cited before building competing workflows that allegedly mirror the patented methods.
What happened: Cleerly issued a statement calling the lawsuit a "lawsuit to limit competition" and said it "strongly disagrees" with the allegations. The company's answer is due July 8. No preliminary injunction motion has been publicly disclosed.
Precedent value: AI-powered medical devices occupy a contested space in patent law: Section 101 eligibility challenges are common, and the specification must tie abstract AI steps to concrete, patent-eligible outputs.
통계 카드를 불러오는 중…
Cases to watch
The following matters from prior weeks have material developments expected in the coming days:
- Sigwalt v. Amazon/Ring (W.D. Wash.): Amazon's answer or motion to dismiss on the BIPA-adjacent biometric class action is due approximately July 1–2. The same week Meta quietly removed facial recognition from its smart glasses — a potentially relevant industry-behavior signal if Amazon argues its feature is standard practice.
- CNN v. Perplexity (S.D.N.Y.): A scheduling order and ruling on CNN's TRO/preliminary injunction motion is expected mid-to-late June.
- Musk / X FTC consent decree petition: The FTC comment period closes July 2, 2026. A ruling is expected shortly after.
- Florida AG v. OpenAI: The state's civil case (filed June 1) is now in the context of the 42-state coalition probe announced this week. OpenAI's motion to dismiss is expected in coming weeks.
참고 출처
- 1OpenAI faces investigation from state attorneys general
- 2ChatGPT Faces 42-State Probe: Sycophancy Design Flaw Named in Subpoena
- 3She confided in ChatGPT the night of her suicide. Now, her mother is suing OpenAI.
- 4Musk's xAI accused of illegally firing engineer who raised safety concerns
- 5Google Challenges Court Ruling Of Liability For False AI-Generated Claims
- 6Google to challenge German ruling saying it is liable for AI-generated false claims
- 7Pixabay — gavel / justice (QuinceCreative)
- 8Patent Infringement Lawsuit Alleges "Piracy" of AI-Driven Medical Technology
이 콘텐츠를 둘러싼 관점이나 맥락을 계속 보강해 보세요.