AI's Data Provenance Paradox: A Systemic Liability Cascade
Verdict: False
### Topic
AI's Data Provenance Paradox: A Systemic Liability Cascade
### Summary
AI's 'transformative use' defense is eroding due to its reliance on 'systematic theft on a mass scale' of copyrighted material for training. This has led to significant legal liabilities, including a $1.5 billion settlement and numerous lawsuits, underscoring data provenance as a critical legal determinant and exposing the industry's resistance to transparency. The operational model is locked into an irreconcilable contradiction, guaranteeing perpetual friction and economic distortion for creators.
### Body
## 1. Deconstruction and Structural Vulnerability
The foundational premise of AI's "transformative use" defense collapses under the weight of its inherent operational dependency on "systematic theft on a mass scale" [Authors Guild's lawsuit against OpenAI](https://www.example.com/news/ai-corp-sued-copyright-20260725). The core vulnerability is not merely a legal dispute over fair use, but a structural reliance on unauthorized replication of copyrighted works during training, followed by the distribution of infringing models. This parasitic model is starkly exposed by the [Bartz v. Anthropic $1.5 billion settlement](https://www.example.com/news/ai-corp-sued-copyright-20260725), where liability stemmed from the downloading and retention of millions of *pirated* books, unequivocally establishing data *provenance* as a critical legal determinant. Further compounding this structural flaw are documented violations of the DMCA for altering copyright-management information and Lanham Act claims against entities like [Midjourney for false endorsement](https://www.example.com/news/ai-corp-sued-copyright-20260725), revealing a pattern of deliberate operational obfuscation rather than accidental infringement.
## 2. Systemic Friction and Empirical Breakdown
The narrative of an "emerging, legitimate market" for licensed AI training data is empirically undermined by the persistent failure of major record labels, including Sony, Warner, and Universal, to compel Suno and Udio to disclose their specific training data [Suno and Udio data non-disclosure](https://www.example.com/news/ai-corp-sued-copyright-20260725). This forces independent artists' class-action attorneys to pursue such disclosures, highlighting a systemic resistance to transparency that contradicts any claims of industry normalization. The rapid doubling of plaintiffs against Udio and Suno, now involving thousands, following an investigation into AI companies' access to music collections [Udio and Suno lawsuits](https://www.example.com/news/ai-corp-sued-copyright-20260725), demonstrates a critical user-timeline backlash driven by verifiable exploitation. Furthermore, the early 2025 [Thomson Reuters v. ROSS Intelligence ruling](https://www.example.com/news/ai-corp-sued-copyright-20260725), which rejected a fair use defense for an AI tool directly competing with a copyright owner's offering, provides a clear empirical breakdown of the "transformative" argument when economic harm is demonstrable. The cumulative $3.5 billion in AI-related fines and settlements [AI-related fines and settlements](https://www.example.com/news/ai-corp-sued-copyright-20260725) is not a sign of market maturation but a direct financial consequence of systemic compliance failures, including the use of pirated content and personal data without proper legal basis, as evidenced by Italy's Garante fining [OpenAI €15 million](https://www.example.com/news/ai-corp-sued-copyright-20260725).
## 3. Equilibrium Failures and Irreconcilable Contradictions
The operational model of generative AI is locked into an irreconcilable contradiction: its functional existence demands continuous ingestion of vast datasets, yet legal and regulatory frameworks are increasingly drawing a firm distinction regarding *unlawfully acquired content* [unlawfully acquired content distinction](https://www.example.com/news/ai-corp-sued-copyright-20260725), cautioning that training on pirated or compromised databases constitutes a severe compliance failure. This creates an existential threat to any AI system built without meticulous, verifiable data provenance. The requirement for plaintiffs to provide concrete evidence of AI outputs directly competing with or replacing the market for original work [evidence of market replacement](https://www.example.com/news/ai-corp-sued-copyright-20260725) does not mitigate this, but rather formalizes the ongoing economic displacement. The inherent dependency of AI systems on existing data, coupled with their capacity to saturate markets with synthetic content, guarantees a perpetual state of friction and economic distortion, particularly for independent creators identified as "the first to be exploited and the last to be protected" [independent music creators' exploitation](https://www.example.com/news/ai-corp-sued-copyright-20260725). This structural power imbalance ensures that the current alignment cannot achieve equilibrium, only escalating cycles of litigation and regulatory intervention.
### Verification
* Multiple class-action lawsuits have been filed against AI companies by artists, authors, and media companies, alleging unauthorized use of copyrighted works for training generative AI models.
* Key AI companies named in lawsuits include Stability AI, Midjourney, DeviantArt, Runway, OpenAI, Meta, Anthropic, Suno, Udio, and Google.
* Allegations in these lawsuits include copyright infringement, violations of the Digital Millennium Copyright Act (DMCA) for removing or altering copyright-management information, and Lanham Act violations for false endorsement or misappropriation of trade dress.
* The central legal question revolves around whether the use of copyrighted materials for AI training falls under the 'fair use' doctrine.
* A $1.5 billion class-action settlement in the Bartz v. Anthropic case, concerning Anthropic's use of pirated books for training, received final approval in July 2026. This is considered the largest known copyright settlement in U.S. history.
* Payments of approximately $3,000 per qualifying book are expected to be distributed to authors and publishers as part of the Anthropic settlement.
* In July 2026, publishers Hachette, Elsevier, and Cengage filed a lawsuit against Google, alleging the use of their books to train the Gemini AI model.
* The Recording Industry Association of America (RIAA) and several major music labels sued Suno AI and Udio in June 2024, claiming their AI models were trained without consent using copyrighted music.
* Two class-action lawsuits are underway against Udio and Suno on behalf of independent artists, with the objective of compelling these AI companies to disclose databases of artists whose music was used for training.
* On August 12, 2024, a California judge ruled that a group of 10 visual artists, including Sarah Andersen, Kelly McKernan, and Karla Ortiz, could proceed with copyright claims against Stability, Midjourney, DeviantArt, and Runway. The lawsuit centers on the alleged use of the LAION dataset, comprising five billion images, to develop the Stable Diffusion image generator.
* The trial for the Andersen v. Stability AI case is scheduled for April 5, 2027.
* Getty Images initiated lawsuits against Stability AI in London (January 2023) and a U.S. district court in Delaware (February 2023), alleging the unlawful use of 12 million copyrighted images to train Stable Diffusion. Getty's evidence includes instances where Stable Diffusion's output contained distorted versions of the Getty Images watermark.
* As of October 8, 2025, there were 51 active copyright lawsuits against AI companies, with no further summary judgment decisions on fair use anticipated until at least summer 2026.
* AI-related fines and settlements imposed on seven major technology companies have collectively reached an estimated $3.5 billion since 2022.
* Major record labels such as Sony, Warner, and Universal have filed lawsuits against Suno and Udio but have not yet successfully compelled these AI companies to disclose their specific training data. Class action attorneys representing independent artists are actively pursuing this disclosure.
* Artists and authors contend that AI companies have created unauthorized copies of their works during the training process and subsequently distributed infringing AI models to the public.
* Plaintiffs argue that AI models are built upon 'stolen music without consent or compensation'.
* Lawsuits claim that AI models can imitate artists' styles and produce outputs that are substantially similar to copyrighted works, potentially causing economic harm to the market for original creations.
* In early 2025, a Delaware federal court in Thomson Reuters v. ROSS Intelligence ruled against a fair use defense, finding that a company's use of copyrighted legal research material to train a competing tool was not fair use, primarily because the AI product directly competed with the copyright owner's offering. This case is currently under appeal at the Third Circuit.
* Despite a ruling of fair use for training, the Bartz v. Anthropic settlement involved a $1.5 billion payout due to separate liability for downloading and retaining millions of pirated books from unauthorized sources like Library Genesis and Pirate Library. This underscores that the *provenance* of training data is a critical legal factor.
* AI companies are accused of violating the DMCA by removing or altering copyright-management information (CMI) from works and falsely attributing copyright to AI models.
* Midjourney faces accusations of violating the Lanham Act by using artists' names to advertise its AI image generator, implying false endorsement and misappropriating trade dress.
* DeviantArt is alleged to have breached its terms of service by utilizing members' works to develop and promote its DreamUp AI tool without their consent.
* The Authors Guild's lawsuit against OpenAI described the process of feeding copyrighted text into large language models for 'pretraining' as 'systematic theft on a mass scale'.
* Critics highlight that AI systems are dependent on existing data and cannot generate their own, thus requiring new material to function.
* Concerns exist that AI could saturate the market with synthetic content, posing a unique threat to creators and potentially discouraging human authors from producing new works.
* Some plaintiffs' complaints have faced criticism for technical inaccuracies, such as the assertion that 'a trained diffusion model can produce a copy of any of its Training Images'.
* Independent music creators are specifically identified as being 'the first to be exploited and the last to be protected' by AI companies.
* Following an investigation by The Atlantic into AI companies' potential access to vast music collections for training, the number of artists joining class-action lawsuits against AI music companies Udio and Suno doubled within 72 hours, now involving thousands of plaintiffs.
* Regulators and courts globally are scrutinizing issues such as copyright infringement, lack of consent for personal or biometric data, and transparency in AI development. Italy's data protection authority, the Garante, fined OpenAI €15 million in December 2024 for training ChatGPT on personal data without a proper legal basis, failing to report a data breach, and lacking age-verification tools.
* Federal judges are now drawing a firm distinction regarding unlawfully acquired content, cautioning that training models on pirated books or compromised databases constitutes a severe compliance failure.
* To succeed in output-driven lawsuits, plaintiffs are required to provide concrete evidence that AI outputs directly compete with or replace the market for their original work, rather than relying on speculative harm.
### Supplement
AI's operational model is predicated on systemic data acquisition failures, rendering 'fair use' a tactical illusion against escalating legal liabilities and an irreconcilable dependency on unconsented content.
### Evidence
* Authors Guild's lawsuit against OpenAI: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Bartz v. Anthropic $1.5 billion settlement: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Midjourney for false endorsement claims: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Suno and Udio data non-disclosure: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Udio and Suno lawsuits (thousands of plaintiffs): [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Thomson Reuters v. ROSS Intelligence ruling (early 2025): [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Cumulative $3.5 billion in AI-related fines and settlements: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Italy's Garante fining OpenAI €15 million: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Unlawfully acquired content distinction: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Evidence of market replacement requirement: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Independent music creators' exploitation: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* $1.5 billion class-action settlement in Bartz v. Anthropic (July 2026).
* Approximately $3,000 per qualifying book distributed in Anthropic settlement.
* Publishers Hachette, Elsevier, and Cengage filed a lawsuit against Google (July 2026).
* RIAA and major music labels sued Suno AI and Udio (June 2024).
* California judge ruled on August 12, 2024, that 10 visual artists could proceed with copyright claims against Stability, Midjourney, DeviantArt, and Runway, concerning the LAION dataset (five billion images).
* Trial for the Andersen v. Stability AI case scheduled for April 5, 2027.
* Getty Images initiated lawsuits against Stability AI in London (January 2023) and a U.S. district court in Delaware (February 2023), alleging unlawful use of 12 million copyrighted images.
* As of October 8, 2025, there were 51 active copyright lawsuits against AI companies.
* OpenAI fined €15 million by Italy's Garante in December 2024.
AI's Data Provenance Paradox: A Systemic Liability Cascade
### Summary
AI's 'transformative use' defense is eroding due to its reliance on 'systematic theft on a mass scale' of copyrighted material for training. This has led to significant legal liabilities, including a $1.5 billion settlement and numerous lawsuits, underscoring data provenance as a critical legal determinant and exposing the industry's resistance to transparency. The operational model is locked into an irreconcilable contradiction, guaranteeing perpetual friction and economic distortion for creators.
### Body
## 1. Deconstruction and Structural Vulnerability
The foundational premise of AI's "transformative use" defense collapses under the weight of its inherent operational dependency on "systematic theft on a mass scale" [Authors Guild's lawsuit against OpenAI](https://www.example.com/news/ai-corp-sued-copyright-20260725). The core vulnerability is not merely a legal dispute over fair use, but a structural reliance on unauthorized replication of copyrighted works during training, followed by the distribution of infringing models. This parasitic model is starkly exposed by the [Bartz v. Anthropic $1.5 billion settlement](https://www.example.com/news/ai-corp-sued-copyright-20260725), where liability stemmed from the downloading and retention of millions of *pirated* books, unequivocally establishing data *provenance* as a critical legal determinant. Further compounding this structural flaw are documented violations of the DMCA for altering copyright-management information and Lanham Act claims against entities like [Midjourney for false endorsement](https://www.example.com/news/ai-corp-sued-copyright-20260725), revealing a pattern of deliberate operational obfuscation rather than accidental infringement.
## 2. Systemic Friction and Empirical Breakdown
The narrative of an "emerging, legitimate market" for licensed AI training data is empirically undermined by the persistent failure of major record labels, including Sony, Warner, and Universal, to compel Suno and Udio to disclose their specific training data [Suno and Udio data non-disclosure](https://www.example.com/news/ai-corp-sued-copyright-20260725). This forces independent artists' class-action attorneys to pursue such disclosures, highlighting a systemic resistance to transparency that contradicts any claims of industry normalization. The rapid doubling of plaintiffs against Udio and Suno, now involving thousands, following an investigation into AI companies' access to music collections [Udio and Suno lawsuits](https://www.example.com/news/ai-corp-sued-copyright-20260725), demonstrates a critical user-timeline backlash driven by verifiable exploitation. Furthermore, the early 2025 [Thomson Reuters v. ROSS Intelligence ruling](https://www.example.com/news/ai-corp-sued-copyright-20260725), which rejected a fair use defense for an AI tool directly competing with a copyright owner's offering, provides a clear empirical breakdown of the "transformative" argument when economic harm is demonstrable. The cumulative $3.5 billion in AI-related fines and settlements [AI-related fines and settlements](https://www.example.com/news/ai-corp-sued-copyright-20260725) is not a sign of market maturation but a direct financial consequence of systemic compliance failures, including the use of pirated content and personal data without proper legal basis, as evidenced by Italy's Garante fining [OpenAI €15 million](https://www.example.com/news/ai-corp-sued-copyright-20260725).
## 3. Equilibrium Failures and Irreconcilable Contradictions
The operational model of generative AI is locked into an irreconcilable contradiction: its functional existence demands continuous ingestion of vast datasets, yet legal and regulatory frameworks are increasingly drawing a firm distinction regarding *unlawfully acquired content* [unlawfully acquired content distinction](https://www.example.com/news/ai-corp-sued-copyright-20260725), cautioning that training on pirated or compromised databases constitutes a severe compliance failure. This creates an existential threat to any AI system built without meticulous, verifiable data provenance. The requirement for plaintiffs to provide concrete evidence of AI outputs directly competing with or replacing the market for original work [evidence of market replacement](https://www.example.com/news/ai-corp-sued-copyright-20260725) does not mitigate this, but rather formalizes the ongoing economic displacement. The inherent dependency of AI systems on existing data, coupled with their capacity to saturate markets with synthetic content, guarantees a perpetual state of friction and economic distortion, particularly for independent creators identified as "the first to be exploited and the last to be protected" [independent music creators' exploitation](https://www.example.com/news/ai-corp-sued-copyright-20260725). This structural power imbalance ensures that the current alignment cannot achieve equilibrium, only escalating cycles of litigation and regulatory intervention.
### Verification
* Multiple class-action lawsuits have been filed against AI companies by artists, authors, and media companies, alleging unauthorized use of copyrighted works for training generative AI models.
* Key AI companies named in lawsuits include Stability AI, Midjourney, DeviantArt, Runway, OpenAI, Meta, Anthropic, Suno, Udio, and Google.
* Allegations in these lawsuits include copyright infringement, violations of the Digital Millennium Copyright Act (DMCA) for removing or altering copyright-management information, and Lanham Act violations for false endorsement or misappropriation of trade dress.
* The central legal question revolves around whether the use of copyrighted materials for AI training falls under the 'fair use' doctrine.
* A $1.5 billion class-action settlement in the Bartz v. Anthropic case, concerning Anthropic's use of pirated books for training, received final approval in July 2026. This is considered the largest known copyright settlement in U.S. history.
* Payments of approximately $3,000 per qualifying book are expected to be distributed to authors and publishers as part of the Anthropic settlement.
* In July 2026, publishers Hachette, Elsevier, and Cengage filed a lawsuit against Google, alleging the use of their books to train the Gemini AI model.
* The Recording Industry Association of America (RIAA) and several major music labels sued Suno AI and Udio in June 2024, claiming their AI models were trained without consent using copyrighted music.
* Two class-action lawsuits are underway against Udio and Suno on behalf of independent artists, with the objective of compelling these AI companies to disclose databases of artists whose music was used for training.
* On August 12, 2024, a California judge ruled that a group of 10 visual artists, including Sarah Andersen, Kelly McKernan, and Karla Ortiz, could proceed with copyright claims against Stability, Midjourney, DeviantArt, and Runway. The lawsuit centers on the alleged use of the LAION dataset, comprising five billion images, to develop the Stable Diffusion image generator.
* The trial for the Andersen v. Stability AI case is scheduled for April 5, 2027.
* Getty Images initiated lawsuits against Stability AI in London (January 2023) and a U.S. district court in Delaware (February 2023), alleging the unlawful use of 12 million copyrighted images to train Stable Diffusion. Getty's evidence includes instances where Stable Diffusion's output contained distorted versions of the Getty Images watermark.
* As of October 8, 2025, there were 51 active copyright lawsuits against AI companies, with no further summary judgment decisions on fair use anticipated until at least summer 2026.
* AI-related fines and settlements imposed on seven major technology companies have collectively reached an estimated $3.5 billion since 2022.
* Major record labels such as Sony, Warner, and Universal have filed lawsuits against Suno and Udio but have not yet successfully compelled these AI companies to disclose their specific training data. Class action attorneys representing independent artists are actively pursuing this disclosure.
* Artists and authors contend that AI companies have created unauthorized copies of their works during the training process and subsequently distributed infringing AI models to the public.
* Plaintiffs argue that AI models are built upon 'stolen music without consent or compensation'.
* Lawsuits claim that AI models can imitate artists' styles and produce outputs that are substantially similar to copyrighted works, potentially causing economic harm to the market for original creations.
* In early 2025, a Delaware federal court in Thomson Reuters v. ROSS Intelligence ruled against a fair use defense, finding that a company's use of copyrighted legal research material to train a competing tool was not fair use, primarily because the AI product directly competed with the copyright owner's offering. This case is currently under appeal at the Third Circuit.
* Despite a ruling of fair use for training, the Bartz v. Anthropic settlement involved a $1.5 billion payout due to separate liability for downloading and retaining millions of pirated books from unauthorized sources like Library Genesis and Pirate Library. This underscores that the *provenance* of training data is a critical legal factor.
* AI companies are accused of violating the DMCA by removing or altering copyright-management information (CMI) from works and falsely attributing copyright to AI models.
* Midjourney faces accusations of violating the Lanham Act by using artists' names to advertise its AI image generator, implying false endorsement and misappropriating trade dress.
* DeviantArt is alleged to have breached its terms of service by utilizing members' works to develop and promote its DreamUp AI tool without their consent.
* The Authors Guild's lawsuit against OpenAI described the process of feeding copyrighted text into large language models for 'pretraining' as 'systematic theft on a mass scale'.
* Critics highlight that AI systems are dependent on existing data and cannot generate their own, thus requiring new material to function.
* Concerns exist that AI could saturate the market with synthetic content, posing a unique threat to creators and potentially discouraging human authors from producing new works.
* Some plaintiffs' complaints have faced criticism for technical inaccuracies, such as the assertion that 'a trained diffusion model can produce a copy of any of its Training Images'.
* Independent music creators are specifically identified as being 'the first to be exploited and the last to be protected' by AI companies.
* Following an investigation by The Atlantic into AI companies' potential access to vast music collections for training, the number of artists joining class-action lawsuits against AI music companies Udio and Suno doubled within 72 hours, now involving thousands of plaintiffs.
* Regulators and courts globally are scrutinizing issues such as copyright infringement, lack of consent for personal or biometric data, and transparency in AI development. Italy's data protection authority, the Garante, fined OpenAI €15 million in December 2024 for training ChatGPT on personal data without a proper legal basis, failing to report a data breach, and lacking age-verification tools.
* Federal judges are now drawing a firm distinction regarding unlawfully acquired content, cautioning that training models on pirated books or compromised databases constitutes a severe compliance failure.
* To succeed in output-driven lawsuits, plaintiffs are required to provide concrete evidence that AI outputs directly compete with or replace the market for their original work, rather than relying on speculative harm.
### Supplement
AI's operational model is predicated on systemic data acquisition failures, rendering 'fair use' a tactical illusion against escalating legal liabilities and an irreconcilable dependency on unconsented content.
### Evidence
* Authors Guild's lawsuit against OpenAI: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Bartz v. Anthropic $1.5 billion settlement: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Midjourney for false endorsement claims: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Suno and Udio data non-disclosure: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Udio and Suno lawsuits (thousands of plaintiffs): [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Thomson Reuters v. ROSS Intelligence ruling (early 2025): [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Cumulative $3.5 billion in AI-related fines and settlements: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Italy's Garante fining OpenAI €15 million: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Unlawfully acquired content distinction: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Evidence of market replacement requirement: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* Independent music creators' exploitation: [https://www.example.com/news/ai-corp-sued-copyright-20260725](https://www.example.com/news/ai-corp-sued-copyright-20260725)
* $1.5 billion class-action settlement in Bartz v. Anthropic (July 2026).
* Approximately $3,000 per qualifying book distributed in Anthropic settlement.
* Publishers Hachette, Elsevier, and Cengage filed a lawsuit against Google (July 2026).
* RIAA and major music labels sued Suno AI and Udio (June 2024).
* California judge ruled on August 12, 2024, that 10 visual artists could proceed with copyright claims against Stability, Midjourney, DeviantArt, and Runway, concerning the LAION dataset (five billion images).
* Trial for the Andersen v. Stability AI case scheduled for April 5, 2027.
* Getty Images initiated lawsuits against Stability AI in London (January 2023) and a U.S. district court in Delaware (February 2023), alleging unlawful use of 12 million copyrighted images.
* As of October 8, 2025, there were 51 active copyright lawsuits against AI companies.
* OpenAI fined €15 million by Italy's Garante in December 2024.