August 2, 2026  ·  Complaint filed July 31, 2026  ·  ilnd, No. 1:26-cv-09153

Newman News, By David Brunk

Read the filing

Berliner v. Huang Accuses NVIDIA Board of Hiding AI Training on Pirated Books and YouTube Videos

Shareholder Says Directors Knew of Copyright Infringement and BIPA Violations While Selling Stock

A verified stockholder derivative complaint filed July 31, 2026, alleges that NVIDIA Corporation’s board of directors and senior executives breached their fiduciary duties by adopting an “ask forgiveness, not approval” strategy to train the company’s artificial-intelligence models on massive datasets of pirated books, copyrighted YouTube videos, and biometric voiceprints without consent. The suit, Jessica Berliner v. Jen-Hsun Huang, et al., names CEO Jensen Huang and 14 other directors and officers as defendants and seeks to hold them personally liable for what the complaint describes as a strategy that exposed NVIDIA to significant legal risks while misleading investors. The complaint alleges violations of Section 14(a) of the Securities Exchange Act and SEC Rule 14a-9 (15 U.S.C. § 78n(a); 17 C.F.R. § 240.14a-9) against the director defendants for issuing false and misleading proxy statements that omitted material facts about NVIDIA’s use of pirated datasets and BIPA violations, thereby influencing stockholder votes (Compl. ¶52). The complaint further alleges that the defendants breached duties of loyalty, care, and good faith by adopting an unlawful business strategy using copyrighted and unlicensed materials for AI development, violating privacy laws such as BIPA, and making false and misleading public statements, in violation of Delaware law governing fiduciary duties (Compl. ¶51). The complaint specifically alleges that the defendants caused NVIDIA to violate copyright laws, federal securities laws, and the Illinois Biometric Information Privacy Act (BIPA) by failing to obtain consent, provide notice, or secure written releases for voiceprints used in commercial voice models (Compl. ¶6, ¶110).

The plaintiff, a stockholder of record, alleges that the defendants knew or recklessly disregarded that NVIDIA’s AI training practices violated copyright law and the Illinois Biometric Information Privacy Act, yet failed to disclose these risks in SEC filings and proxy statements. The complaint ties the alleged misconduct to a stock-repurchase program and stock sales by Huang during the relevant period. The complaint further alleges that the director defendants made false and misleading statements and omissions in SEC filings, proxy statements, and public disclosures about NVIDIA’s AI training practices, which artificially inflated the company’s stock prices (Compl. ¶53). The complaint states that NVIDIA’s unlawful business strategy involved using "copyrighted and unlicensed materials" and that the defendants followed an "ask forgiveness not approval" model, as quoted in the filing: "Defendants followed the 'ask forgiveness not approval' model and adopted and implemented an unlawful plan to use a pirated dataset..." (Compl. ¶7). The complaint also alleges that the defendants caused NVIDIA to repurchase $13.258 billion of its own stock at artificially inflated prices due to these false and misleading statements (Compl. ¶53).

NVIDIA’s AI Models Trained on Pirated Books, Complaint Alleges

The complaint alleges that NVIDIA trained its large-language models on a dataset of pirated books sourced from shadow libraries, including The Pile, which contained the Books3 dataset of approximately 196,640 pirated books (Compl. ¶3). Internal documents cited in the filing show that NVIDIA turned to pirated sources after publishers rejected licensing deals. The complaint describes how NVIDIA management allegedly approved the use of pirated materials from one such source within a week of contacting the site. The filing quotes the source as stating that large-language models thrive on high-quality data and that it possessed the largest collection of books, papers, and magazines in the world. The complaint adds that major companies building such models contacted this source to train on its data, and that NVIDIA was particularly focused on book datasets, knowing that published books under copyright were the most valuable for developing these models (Compl. ¶71-73). The dataset was removed in October 2023 due to copyright-infringement claims, but not before NVIDIA had incorporated it into its training pipelines. The complaint alleges that NVIDIA unlawfully copied copyrighted works from shadow libraries including Bibliotek, Anna’s Archive, LibGen, Sci-Hub, and Z-Library (Compl. ¶4). The complaint specifically alleges that Bibliotek is "well known as a notorious pirate website that contains vast quantities of unauthorized copyright materials..." (Compl. ¶62).

The complaint details NVIDIA’s compensation figures for its executives, which the plaintiff alleges were excessive given the company’s alleged unlawful practices. Colette M. Kress, Executive Vice President and CFO, received $14.4 million in 2025 compensation, $21.4 million in 2024 compensation, and $13.3 million in 2023 compensation (Compl. ¶ amounts). Ajay K. Puri, Executive Vice President of Worldwide Field Operations, received $14.8 million in 2025 compensation, $21.6 million in 2024 compensation, and $13.6 million in 2023 compensation (Compl. ¶ amounts). Deborah Shoquist, Executive Vice President of Operations, received $14.3 million in 2025 compensation, $19.2 million in 2024 compensation, and $11.1 million in 2023 compensation (Compl. ¶ amounts). The complaint alleges that these compensation figures, along with stock sales by the defendants, constitute unjust enrichment while NVIDIA’s financial performance was artificially inflated by the unlawful practices (Compl. ¶53). The complaint further alleges that the defendants received excessive compensation and profits from stock sales during this period, including board-related compensation of $363,809 for multiple defendants in 2025 (Compl. ¶ amounts).

In August 2023, the NVIDIA Board authorized a $25 billion share repurchase program, which the complaint alleges was conducted while the company’s stock price was artificially inflated due to the defendants’ false and misleading statements (Compl. ¶ timeline). This program was followed by an even larger authorization in August 2024, when the Board approved a $50 billion share repurchase program (Compl. ¶ timeline). The complaint alleges that these repurchase programs were part of a broader strategy to waste corporate assets by buying back stock at inflated prices, including the repurchase of 7.3 million shares at an average price of $186.52 between December 22, 2025, and January 25, 2026 (Compl. ¶ timeline).

NVIDIA Scraped Millions of YouTube Videos, Suit Says

The complaint alleges that NVIDIA scraped and downloaded millions of copyrighted videos from YouTube to train its foundational video world model used in products including robotics, avatar technology, simulation platforms, and gaming. The complaint states that NVIDIA’s actions constitute a violation of the Digital Millennium Copyright Act (DMCA), 17 U.S.C. § 1201(a), by circumventing YouTube’s technological protection measures to access and use these videos (Compl. ¶127). The complaint alleges that NVIDIA trained AI models on copyrighted works, including books, videos, and voiceprints, without consent, credit, or compensation, in violation of 17 U.S.C. §§ 101 et seq. (Compl. ¶54). The complaint further alleges that NVIDIA improperly accessed the actual audio and video files from YouTube and stored them indefinitely in its own system, stating, "NVIDIA 'improperly accessed the actual audio and video files' from YouTube, storing them indefinitely in its own system" (Compl. ¶53).

These datasets included tens of millions of video clips extracted from YouTube content. The HD-VG-130M dataset alone contained 1,549,408 YouTube videos, resulting in 130 million clips, while the HDVILA-100M dataset included 3,098,462 YouTube videos, producing 100 million clips (Compl. ¶78). The HowTo100M dataset comprised 1,238,911 YouTube videos, yielding another 100 million clips. The complaint alleges that NVIDIA circumvented YouTube’s technological protection measures to access and use these videos. The filing states that bulk extraction of YouTube videos cannot occur without bypassing these protections, quoting an internal communication: "Bulk extraction of YouTube videos cannot occur without circumventing YouTube’s TPMs" (Compl. ¶91). The Panda-70M dataset, compiled from 3.8 million high-resolution videos, resulted in 70.8 million clips and was trained on 48 NVIDIA A100 GPUs (Compl. ¶81).

NVIDIA employees allegedly used tools such as “yt-dlp” to avoid detection and acknowledged in internal communications that their actions violated YouTube’s terms of service. One employee noted that every retrieval required circumventing YouTube’s restrictions and licensing limits, stating, “every retrieval requires the user to circumvent YouTube’s technological restrictions, terms of service, and licensing limits” (Compl. ¶77). Another employee highlighted missing videos in one of the datasets, noting, “Found a list of files to download with high priority. Turns out there are -2.3M raw videos missing in the HDVILA dataset we have!” (Compl. ¶83).

NVIDIA employees raised legal and ethical concerns about the YouTube scraping in internal messages. One employee asked whether using copyrighted material was permissible, and another responded that while the legal status was unclear, the company’s legal team had approved similar practices for other types of training, stating, “whether using copyrighted material is fair use is currently an open legal issue [but] our legal team has OK’ed this kind of thing for LLM training” (Compl. ¶96). A third employee questioned whether the legal team had approved a specific dataset download, and a senior researcher, Ming-Yu Liu, replied that it was an executive decision and that there was umbrella approval for all data, stating, “This is an executive decision. We have umbrella approval of all the data” (Compl. ¶97). The complaint alleges that NVIDIA treated these datasets as “raw materials” for its Cosmos AI model, despite most containing licensed material intended for non-commercial or academic use (Compl. ¶76). The complaint further alleges that NVIDIA’s use of these datasets was part of a broader strategy to generate a "huge curated video dataset for video generative modeling," as described in internal documents (Compl. ¶76).

Voice Synthesis Models Trained on Hundreds of Thousands of Hours of Speech Without Consent, Suit Alleges

The complaint alleges that NVIDIA trained its voice synthesis models on hundreds of thousands of hours of human speech without obtaining consent, notice, or written releases from the speakers, in violation of the Illinois Biometric Information Privacy Act (BIPA) (740 ILCS 14/1 et seq.) (Compl. ¶110). The filing states that NVIDIA’s models were trained on voiceprints extracted without the knowledge or consent of the speakers, and that NVIDIA failed to identify the source speakers, provide written notice of the specific purpose and duration of collection, or obtain a written release, as required by BIPA (Compl. ¶6). The complaint alleges that NVIDIA failed to obtain consent, provide notice, or secure written releases for voiceprints used in commercial voice models (Compl. ¶110). The complaint further alleges that NVIDIA chose speed and scale over compliance, stating, "NVIDIA chose speed and scale over compliance" (Compl. ¶128).

The filing details datasets allegedly used by NVIDIA, including one containing a million hours of speech (Granary dataset) and another with 70,000 hours, with no evidence of proper licensing or consent. The complaint alleges that NVIDIA obtained substantial portions of its training data from publicly accessible internet sources without the consent of source speakers. Voiceprints, described as biometric identifiers, were encoded into commercial models and distributed via open-weight releases such as PersonaPlex, Canary, and Parakeet. The complaint quotes documentation for one of NVIDIA’s models as stating, “The Magpie TTS Zeroshot model takes the text and audio prompt as input, then generates the corresponding audio of the input text with the target speaker’s voice” (Compl. ¶104). The complaint further alleges that NVIDIA’s foundational voice synthesis models were trained on voiceprints extracted without the knowledge or consent of the speakers, stating, “NVIDIA’s foundational voice synthesis models were trained on voiceprints extracted without the knowledge or consent of the speakers” (Compl. ¶107). The complaint quotes a WaveGlow paper as stating, “Speech synthesis requires generating very high dimensional samples with strong long term dependencies” (Compl. ¶101). The complaint also alleges that NVIDIA’s voice synthesis pipeline creates biometric identifiers, which are subject to BIPA liability, and that the company distributed these models without complying with the law’s requirements (Compl. ¶105).

Defendants Knew of Litigation Risks, Complaint Alleges

The complaint alleges that the defendants were aware of the legal risks posed by NVIDIA’s AI training practices but failed to disclose them to investors. The filing cites damages paid by another AI developer, Anthropic, which paid $1.5 billion to resolve a copyright-infringement action, as a signal of potential liability for NVIDIA (Compl. ¶12). The complaint also references six related lawsuits filed against NVIDIA between 2024 and 2026, including claims of copyright infringement and violations of privacy laws. These include Nazemian v. NVIDIA Corp. (copyright infringement), Ted Entertainment, Inc. v. NVIDIA Corp. (copyright infringement), Rogers v. NVIDIA Corp. (BIPA violations), Youngblood v. NVIDIA Corp. (copyright infringement), Beaulier v. NVIDIA Corp. (copyright infringement), and S.A. Jamendo v. NVIDIA Corp. (copyright infringement, breach of contract, unjust enrichment, and unfair business practices) (Compl. ¶121-122). On May 5, 2026, the court denied in part NVIDIA’s motion to dismiss in the Nazemian action, and the company’s stock dropped to $196.50 following the ruling (Compl. ¶ timeline). By June 26, 2026, NVIDIA’s stock closed at $192.53 (Compl. ¶ timeline).

The complaint alleges that NVIDIA’s internal documents show the company turned to piracy due to competitive pressures and the inability to secure licensed books. The filing states that the defendants adopted and implemented a plan to use datasets to train NVIDIA’s AI models, quoting an internal source as stating, “It is well understood that LLMs thrive on high-quality data. We have the largest collection of books, papers, magazines, etc. in the world... Then came AI. Virtually all major companies building LLMs contacted us to train on our data” (Compl. ¶71-72). The complaint further alleges that NVIDIA was “hyper [f]ocused on books corpuses” and knew that “published books under copyright” were “the most valuable” for developing large-language models (Compl. ¶73). The complaint quotes the company’s code of conduct as requiring officers and directors to put the company’s interests ahead of their own, stating, “put the Company interests ahead of their own” (Compl. ¶45), but alleges that the defendants breached this duty by prioritizing speed and scale over compliance.

The complaint quotes a company code of conduct requiring full, fair, accurate, timely, and clear disclosures, stating, “We’re committed to full, fair, accurate, timely, and clear disclosures in reports and documents” (Compl. ¶47(b)), but alleges that NVIDIA’s SEC filings omitted material facts about its use of datasets and privacy-law violations. The complaint specifically alleges that NVIDIA’s 2023 Annual Report contained false and misleading statements by omitting the company’s use of massive datasets obtained by circumventing access controls, such as those from YouTube and pirated libraries (Compl. ¶137). The complaint alleges that the defendants caused NVIDIA to violate copyright laws, federal securities laws, and BIPA, and made improper public statements and failed to implement adequate internal controls (Compl. ¶52). The complaint further alleges that the defendants breached their duty to prevent the dissemination of inaccurate and untruthful information about NVIDIA’s financial condition, operations, and internal controls, stating, "the defendants had a duty to prevent and not to effect the dissemination of inaccurate and untruthful information" (Compl. ¶39).

On May 20, 2026, NVIDIA announced record Q1 FY 2027 earnings of $81.6 billion, up 20% from the previous quarter and 85% from the previous year, and announced an $80 billion additional share repurchase authorization (Compl. timeline). The complaint alleges that this announcement, along with the company’s financial disclosures, contained material omissions regarding the legal risks associated with its AI training practices. The complaint further alleges that NVIDIA’s 2024 Annual Report contained false representations and material omissions regarding the company’s AI platform products, which were signed by the director defendants, Colette M. Kress, and Timothy S. Teter (Compl. ¶138).

Jensen Huang Sold Stock While Aware of Material Risks, Suit Alleges

The complaint alleges that Jensen Huang sold NVIDIA stock between July and October 2025 while in possession of material, non-public information about the company’s copyright-infringement and BIPA violations, in violation of Section 20A of the Securities Exchange Act (15 U.S.C. § 78t-1) (Compl. ¶53). Huang’s total stock sales since June 2025 exceeded $2.9 billion, including sales of 225,000 shares for proceeds of $36.4 million on July 12, 2025, and over $1 billion in proceeds from sales of 5,250,000 shares between July 15, 2025, and October 29, 2025 (Compl. ¶117). On June 7, 2024, NVIDIA’s stock price reached $1,208.90, but following the 10-for-1 stock split on June 10, 2024, the stock price deflated to $121.79 (Compl. ¶ timeline).

Huang’s compensation for 2025 was $36.3 million, down from $49.9 million in 2024 and $34.2 million in 2023. He held over 851,983,603 shares of NVIDIA stock as of the complaint’s filing, representing approximately 3.5% of the company and valued at over $173.9 billion (Compl. ¶44). The complaint alleges that the defendants received excessive compensation and profits from stock sales while NVIDIA’s financial performance was artificially inflated by the unlawful practices (Compl. ¶53). The complaint cites board-related compensation of $363,809 for multiple defendants in 2025, in addition to the individual compensation figures for Huang, Kress, Puri, and Shoquist (Compl. ¶ amounts). The complaint further alleges that the defendants breached their fiduciary duties by failing to exercise the highest obligation of fair dealing, as required by Delaware law, stating, "the highest obligation of fair dealing" (Compl. ¶36).

Complaint Alleges False and Misleading Statements in SEC Filings

The complaint alleges that NVIDIA’s SEC filings, including its annual reports and proxy statements, contained false and misleading statements and material omissions regarding the company’s AI training practices. One annual report stated that innovation was at the company’s core and highlighted over $37 billion in research and development investments since NVIDIA’s inception but omitted details about the datasets used for AI training, stating, “Innovation is at our core. We have invested over $37 billion in research and development since our inception...” (Compl. ¶137). The complaint alleges that the 2024 Annual Report contained false representations and material omissions regarding NVIDIA’s AI platform products, which were signed by the director defendants, Colette M. Kress, and Timothy S. Teter (Compl. ¶138). The complaint further alleges that NVIDIA’s accelerated computing platform leverages GPUs, CUDA, and networking technologies to deliver order-of-magnitude performance advantages over legacy approaches in Data Center, Gaming, Professional Visualization, and Automotive markets, but failed to disclose the legal risks associated with its AI training practices (Compl. ¶58-59). The complaint states that NVIDIA’s unified architecture supports multiple multi-billion-dollar end markets through shared underlying technology and software stacks, but omits any mention of the company’s use of pirated datasets or BIPA violations (Compl. ¶59).

The complaint alleges that NVIDIA’s 2023 Proxy Statement, filed with the SEC on May 2023, contained material omissions regarding the company’s use of pirated datasets and BIPA violations. The complaint also references the 2024 Proxy Statement, filed on May 14, 2024, and the 2025 Annual Report, filed on February 26, 2025, as containing similar omissions (Compl. timeline). The complaint alleges that the director defendants issued false and misleading proxy statements that omitted material facts about NVIDIA’s use of pirated datasets and BIPA violations, influencing stockholder votes (Compl. ¶52). The complaint further alleges that the defendants breached their duty to exercise reasonable supervision over NVIDIA’s financial affairs, as required by Delaware law (Compl. ¶40).

Derivative Claims Seek to Hold Directors and Officers Personally Liable

The complaint asserts derivative claims on behalf of NVIDIA, seeking to hold the defendants personally liable for breaches of fiduciary duty, securities-law violations, waste of corporate assets, and unjust enrichment. The claims include:

• Count I: Breach of Fiduciary Duty: The complaint alleges that the defendants breached their duties of loyalty, care, and good faith by adopting a strategy that violated copyright and privacy laws, and making false and misleading public statements. The filing states that the defendants owed the company and its stockholders fiduciary obligations of care and loyalty, including the highest obligation of fair dealing and a duty to prevent the dissemination of inaccurate information, stating, “each of the Defendants owed and owe NVIDIA and its stockholders fiduciary obligations of care and loyalty” and “the highest obligation of fair dealing” (Compl. ¶35-36). The complaint alleges that the defendants breached these duties by causing NVIDIA to violate copyright laws, federal securities laws, and BIPA, and by making improper public statements and failing to implement adequate internal controls (Compl. ¶52). The complaint further alleges that the defendants exercised control over the alleged wrongful acts and that their conduct involved knowing or culpable violations of their duties, absence of good faith, or reckless disregard for their obligations (Compl. ¶38). The complaint states that the defendants had a duty to exercise reasonable supervision over NVIDIA’s financial affairs and to prevent the dissemination of inaccurate and untruthful information about the company’s financial condition, operations, and internal controls (Compl. ¶39-40). The complaint quotes the company’s code of conduct as requiring officers and directors to put the company’s interests ahead of their own, stating, “put the Company interests ahead of their own” (Compl. ¶45). The complaint also alleges that the defendants breached their duty to act in good faith, stating, "the defendants breached their duties as officers and directors, demonstrating a lack of good faith and reckless disregard for their obligations" (Compl. ¶51).
• Count II: Violations of Section 14(a) of the Securities Exchange Act and SEC Rule 14a-9: The complaint alleges that the director defendants issued proxy statements that omitted material facts about NVIDIA’s use of pirated datasets and BIPA violations, influencing stockholder votes, in violation of 15 U.S.C. § 78n(a) and 17 C.F.R. § 240.14a-9 (Compl. ¶52). The complaint further alleges that these proxy statements contained false and misleading statements that failed to disclose the company’s unlawful business strategy, stating, "Defendants followed the 'ask forgiveness not approval' model and adopted and implemented an unlawful plan to use a pirated dataset..." (Compl. ¶7).
• Count III: Violations of Section 10(b) of the Securities Exchange Act and SEC Rule 10b-5: The complaint alleges that the director defendants made false and misleading statements and omissions in SEC filings, proxy statements, and public disclosures about NVIDIA’s AI training practices, which artificially inflated the company’s stock prices, in violation of 15 U.S.C. § 78j(b) and 17 C.F.R. § 240.10b-5 (Compl. ¶53). The complaint alleges that these statements included claims about the company’s commitment to innovation and ethical standards, while omitting material facts about its use of pirated datasets and BIPA violations. The complaint quotes the company’s code of conduct as stating, "We’re committed to full, fair, accurate, timely, and clear disclosures in reports and documents" (Compl. ¶47(b)), but alleges that the defendants failed to uphold this commitment.
• Count IV: Violations of Section 20A of the Securities Exchange Act: The complaint alleges that Jensen Huang sold NVIDIA stock while in possession of material, non-public information about the company’s copyright-infringement and BIPA violations, in violation of 15 U.S.C. § 78t-1 (Compl. ¶53). The complaint alleges that Huang’s stock sales were conducted while he was aware of the company’s unlawful practices and the associated legal risks, including the potential for significant damages and reputational harm.
• Waste of Corporate Assets: The complaint alleges that the defendants caused NVIDIA to repurchase $13.258 billion of its own stock at artificially inflated prices due to false and misleading statements (Compl. ¶53). The repurchases occurred between June 2024 and January 2026, with NVIDIA repurchasing 25.4 million shares at an average price of $142.67 between October 28, 2024, and November 24, 2024; 10.6 million shares at an average price of $136.86 between November 25, 2024, and December 22, 2024; and 19.3 million shares at an average price of $39.30 between December 23, 2024, and January 26, 2025 (Compl. ¶ timeline). The complaint further alleges that NVIDIA repurchased 5.6 million shares at an average price of $198.89 between October 27, 2025, and November 23, 2025, and 6.6 million shares at an average price of $179.41 between November 24, 2025, and December 21, 2025 (Compl. ¶ timeline). Additionally, the complaint alleges that NVIDIA repurchased 7.3 million shares at an average price of $186.52 between December 22, 2025, and January 25, 2026 (Compl. ¶ timeline). The complaint states that the defendants caused NVIDIA to repurchase its stock at artificially inflated prices due to the false and misleading statements about the company’s AI training practices (Compl. ¶53). The complaint further alleges that these repurchases were part of a broader pattern of waste, including the $25 billion share repurchase program authorized in August 2023 and the $50 billion program authorized in August 2024 (Compl. ¶ timeline).
• Unjust Enrichment: The complaint alleges that the defendants received excessive compensation and profits from stock sales while NVIDIA’s financial performance was artificially inflated by the unlawful practices (Compl. ¶53). The complaint cites board-related compensation of $363,809 for multiple defendants in 2025, in addition to the individual compensation figures for Huang, Kress, Puri, and Shoquist (Compl. ¶ amounts). The complaint further alleges that the defendants breached their duty to act in good faith by receiving excessive compensation while the company’s financial performance was artificially inflated, stating, "the defendants received excessive compensation and profits from stock sales while NVIDIA’s financial performance was artificially inflated by the unlawful practices" (Compl. ¶53).

The complaint also references related claims in other litigation, including:

• Count 1: Copyright Infringement (17 U.S.C. §§ 101 et seq.): The complaint alleges that NVIDIA trained its AI models on copyrighted works, including books, videos, and voiceprints, without consent, credit, or compensation (Compl. ¶54). The complaint states that NVIDIA scraped and downloaded millions of copyrighted videos from YouTube to feed, train, improve, and commercialize its AI model named Cosmos, stating, “scraped and downloaded millions of copyrighted videos from ... YouTube, in order to feed, train, improve and commercialize NVIDIA’s AI ‘model named ‘Cosmos.’’” (Compl. ¶1). The complaint further alleges that NVIDIA’s actions constituted an unconscionable attack on the community of content creators, stating, “an unconscionable attack on the community of content creators whose content is used to fuel the multi-trillion-dollar generative AI industry without any compensation” (Compl. ¶52). The complaint alleges that NVIDIA’s use of copyrighted materials was widespread, including the use of the Books3 dataset, which consisted of approximately 196,640 pirated books (Compl. ¶60).
• Count 2: Violation of Digital Millennium Copyright Act (DMCA) (17 U.S.C. § 1201(a)): The complaint alleges that NVIDIA circumvented YouTube’s technological protection measures to scrape copyrighted videos for AI training (Compl. ¶127). The complaint states that NVIDIA improperly accessed the actual audio and video files from YouTube and stored them indefinitely in its own system (Compl. ¶53). The complaint alleges that bulk extraction of YouTube videos cannot occur without circumventing YouTube’s technological protection measures, stating, “Bulk extraction of YouTube videos cannot occur without circumventing YouTube’s TPMs” (Compl. ¶91). The complaint further alleges that NVIDIA used tools such as "yt-dlp" to bypass YouTube’s monitoring systems and avoid detection (Compl. ¶90).
• Count 3: Violation of Illinois Biometric Information Privacy Act (BIPA) (740 ILCS 14/1 et seq.): The complaint alleges that NVIDIA failed to obtain consent, provide notice, or secure written releases for voiceprints used in commercial voice models (Compl. ¶110). The complaint states that NVIDIA ingested hundreds of thousands of hours of human speech to extract voiceprints without complying with BIPA’s requirements, stating, “NVIDIA ingested ‘hundreds of thousands of hours of human speech’ to extract voiceprints without BIPA compliance (notice, consent, or written release)” (Compl. ¶128). The complaint further alleges that NVIDIA chose speed and scale over compliance, stating, “NVIDIA chose speed and scale over compliance” (Compl. ¶128). The complaint alleges that NVIDIA’s voice synthesis models were trained on millions of distinct speakers’ voiceprints without their knowledge or consent (Compl. ¶128).

The complaint seeks damages, equitable relief, and corporate governance reforms, including changes to NVIDIA’s internal controls and disclosure practices.

The allegations described here are taken from the filing and remain unproven; no responsive pleading is reflected in the source document.

David Brunk is a civil litigation attorney. He can be reached at david@newmanbrunk.com.