Glossary

The terms, defined.

Plain English where possible; doctrinal precision where it matters. The vocabulary that comes up across every AI lawsuit.

Definitions for legal and technical terms that come up often in AI litigation. Plain English where possible; doctrinal precision where it matters.

A–C

BIPA: Illinois Biometric Information Privacy Act, 740 ILCS 14. Provides per-violation statutory damages (US$1,000 negligent / US$5,000 intentional) for collection or use of biometric identifiers without informed consent. Largest source of AI-privacy class-action exposure. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
CCPA / CPRA: California Consumer Privacy Act and its 2023 amendment. Adds automated-decisionmaking and training-data disclosure requirements applicable to many AI products. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
Class certification: Court order allowing a lawsuit to proceed on behalf of a defined group. The threshold most AI privacy and copyright cases must clear before settlement leverage shifts to plaintiffs. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
Contributory infringement: Common-law theory of secondary copyright liability where a defendant materially contributes to direct infringement by another with knowledge of the infringing activity. Often pled against model deployers and platform hosts. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →

D–F

DMCA §1202: Provision of the Digital Millennium Copyright Act prohibiting removal or alteration of copyright management information (CMI). Plaintiffs increasingly plead §1202 alongside direct infringement; the live question is whether stripping CMI during training meets the statute’s knowledge and intent thresholds. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
DPA: Data Processing Agreement. The contract required by GDPR Art. 28 between a controller and a processor of personal data; central to vendor-due-diligence in AI training pipelines. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
DSM Directive: EU Directive 2019/790 on copyright in the Digital Single Market. Articles 3 and 4 set the EU’s text-and-data-mining exceptions; Article 4 (commercial TDM) was the central exception litigated in GEMA v. OpenAI. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
Fair dealing (UK): UK statutory defense to copyright infringement (CDPA 1988 ss.29–30A). Narrower than U.S. fair use: limited to specified purposes (research, criticism, quotation, parody, news reporting). The UK does not yet have a general TDM exception equivalent to Art. 4 DSM. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
Fair use: Affirmative defense under 17 U.S.C. §107. The four-factor test (purpose, nature, amount, market effect) is the central battleground in training-data copyright cases. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
FEHA: California Fair Employment and Housing Act. New regulations (2025–26) extend disparate-impact and disclosure obligations to automated employment decision tools. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →

G–L

GDPR Article 22: EU General Data Protection Regulation provision granting data subjects the right not to be subject to solely automated decisions, including profiling. The cleanest doctrinal foothold for European challenges to AI-driven decisions. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
Hallucination: Confidently presented but factually false output from an AI model. Now a litigated theory in defamation (Walters v. OpenAI) and false-light cases. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
LGPD: Lei Geral de Proteção de Dados (Brazil’s general data protection law, Law 13.709/2018). Brazil’s GDPR analogue; relevant for AI products with Brazilian users. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
Local Law 144 (NYC): Requires bias audits of automated employment decision tools used by NYC employers and notice to candidates. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →

M–R

MDL: Multidistrict litigation. Procedural mechanism (28 U.S.C. §1407) consolidating related federal cases before a single judge for pretrial proceedings. The SDNY MDL 25-md-03143 consolidates many of the OpenAI publisher and author actions. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
Memorization: Phenomenon where a trained model can reproduce specific training-corpus content verbatim. Central to output-infringement theories; in GEMA v. OpenAI the Munich court treated memorization in model parameters as itself constituting reproduction. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
Motion to dismiss: Pre-discovery motion arguing the complaint fails to state a legally viable claim. Many AI cases survive this stage; the next pressure point is summary judgment. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
PII: Personally identifiable information. Umbrella term for any data that can identify a natural person, directly or in combination. Used loosely; statutes (CCPA, GDPR “personal data”) define the scope precisely. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
RAG: Retrieval-augmented generation. AI systems that retrieve and surface third-party content alongside model output. New legal pressure point — see Dow Jones v. Perplexity. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →

S–Z

Secondary infringement: Umbrella term for indirect copyright liability theories — contributory, vicarious, and (by analogy) inducement. Increasingly important in AI cases targeting upstream model authors and dataset hosts. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
Section 230: 47 U.S.C. §230. Immunizes online services from liability for third-party content. Whether it shields generative model output is an open question; Garcia v. Character.AI declined to apply §230 to design-defect claims, and the Third Circuit’s Anderson v. TikTok is cited as supporting first-party-expression analysis. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
Statutory damages: Damages set by statute rather than proven actual loss. Under U.S. copyright (17 U.S.C. §504(c)), available for timely-registered works at $750–$30,000 per work, or up to $150,000 per work for willful infringement. Drives much of the headline-figure exposure in AI training-data cases. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
Summary judgment: Pre-trial decision based on undisputed facts. The decisive stage in most AI copyright cases (Bartz, Kadrey, Thomson Reuters v. Ross). Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
TDM exception: Text-and-data-mining exception. EU Articles 3 (research) and 4 (commercial, with rights-holder reservation) of the DSM Directive — the European analogue to fair use for training data, narrowed by the Munich Regional Court’s GEMA v. OpenAI ruling. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
Take It Down Act: 2025 U.S. federal law creating a private right of action for non-consensual deepfakes; reshapes deepfake-tort landscape alongside state right-of-publicity statutes. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
Vicarious infringement: Secondary-liability theory imposing copyright liability on a defendant who has the right and ability to supervise infringing activity and a direct financial interest in it — even without knowledge. Pled against platform hosts and deployers. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →
Willful infringement: Copyright infringement committed with knowledge or reckless disregard of the rights-holder’s rights. Triggers enhanced statutory damages up to $150,000 per work under 17 U.S.C. §504(c)(2). Plaintiffs in AI training-data cases routinely plead willfulness based on internal-document discovery. Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more → Read more →

Last updated April 26, 2026.