The rise of generative AI (GenAI) is challenging a legal and ethical foundation of the open-source software movement: attribution. Open-source licenses, such as Apache 2.0, require developers to credit the original creators, but GenAI models trained on massive code corpora cannot practically trace or attribute their outputs. This “technical impossibility” is being used by AI companies as an excuse to disregard licensing terms, a stance challenged in cases like Doe 1 v. GitHub, where the court allowed license violations to be pursued as contract claims, independent of fair use. The chapter argues that this failure to comply could undermine the collaborative culture of open source. The path forward requires either technical attribution solutions, new licensing frameworks specifically for AI, or legal pressure to force compliance. The core question is whether the open-source revolution will be sacrificed for the sake of the AI revolution.

The rise of generative AI (GenAI) systems trained on content harvested by web-scraping bots threatens the economic viability of the open web. Current legal challenges, often focused on copyright law, fail because the fair use defense allows for uncompensated exploitation. This chapter proposes a new legal framework: reconceptualizing a website – specifically its domain name and underlying architecture – as personal property with a strong right to exclude. This shifts the focus from the content to the container. By arguing that GenAI scraping impairs a website’s value (traffic, ad revenue, subscriber potential) as an economic asset, this approach revitalizes the common law tort of trespass to chattels (Restatement (Second) of Torts § 218). This strategy would empower website owners to enforce exclusionary rights and compel AI companies to negotiate for licenses, ultimately fostering a more sustainable and equitable digital ecosystem.

This chapter introduces GenAIuflecting: the reflexive, undemocratic ceding of power and policy to AI companies based on extraordinary, often speculative claims. It argues that this deference is a pattern – from the government uncritically accepting fictional job creation figures to granting major contracts and privileges, such as offering an enterprise AI system to the entire US government for one dollar per agency. At the core of this pattern is AI illiteracy among policymakers and the resulting belief in GenAI exceptionalism, the idea that generative AI (GenAI) deserves special treatment not granted to other industries or even human creators.

The text details how GenAIuflecting manifests across four critical domains: environmental, safety, security, and intellectual property. The chapter concludes by highlighting the dangerous self-reinforcing cycle of concessions, warning that this dynamic is not inevitable but a set of choices serving powerful interests at the expense of workers, creators, communities, and citizens’ safety.

This chapter provides a high-level overview of the complete development and deployment process for generative AI (GenAI) language models, such as those that power systems like ChatGPT and Gemini. It begins by clarifying the distinction between a GenAI model (the core algorithm and code) and a GenAI system (the model plus all necessary infrastructure, user interface, and data management). The discussion then demystifies the technology. Large language models (LLMs) operate by calculating the statistical probability of the next word in a sequence – making them sophisticated, yet fundamentally logic-based, “next-token predictor machines” that lack true comprehension or inherent powers of judgment.

This chapter explores the escalating legal battles that could force AI companies to pay billions in damages and radically restructure how their models are built. At the heart of the dispute is the use of vast troves of copyrighted material, often pirated from sources like LibGen, to train powerful AI models.

The primary legal defense for generative AI (GenAI) companies against copyright infringement claims is fair use. This defense revolves around four key factors, with the concepts of transformative use and the effect on the potential market being the most critical. The outcome of this litigation will determine the future landscape of both AI development and the creative industries. If courts rule broadly for fair use, AI companies will maintain free rein; if they rule against it, the industry will be compelled to adopt licensing or new, noncopying technical approaches. Ultimately, a legal framework conceived in the era of the printing press must now establish the ground rules for the relationship between human and machine creativity in the digital age.

This chapter explores the legal crisis at the intersection of contract law and generative AI (GenAI). Millions of users unknowingly surrender rights to their digital content through “clickwrap” and “browsewrap” agreements they never read, creating a powerful “legal fiction” of consent. GenAI companies are now leveraging these historical agreements to monetize vast repositories of user-generated content, as exemplified by the $60 million Reddit/Google licensing deal. The central legal puzzle is whether the same terms of service (ToS) that bind human users also apply to automated scraping bots that harvest data for AI training. The answer will determine who controls and profits from human creativity in the digital age, either by creating a massive legal loophole for AI companies or by forcing them to negotiate compensation with content creators.

This chapter challenges the assumption that placing a “human in the loop” is an adequate safeguard against the harms of AI. It introduces the concept of automation bias – a cognitive tendency for human decision-makers to over-rely on automated systems, often ignoring contradictory evidence or their own training. The chapter demonstrates how the human operator can become a mere “rubber stamp” for a flawed machine.

The risk is magnified with generative AI (GenAI), whose polished, authoritative-sounding outputs and ubiquitous availability deepen trust and erode critical thinking, leading to both errors of commission (following bad advice) and errors of omission (failing to monitor for errors). The chapter concludes that effective mitigation requires more than just human oversight; it demands intentional changes to system design, organizational culture, and human training to foster a healthy skepticism toward AI recommendations.

This chapter serves as a primer for understanding the sweeping implications of generative AI (GenAI). It begins by establishing a clear terminological foundation, distinguishing modern GenAI from its predecessors, good old-fashioned AI (GOFAI) and machine learning. While GOFAI and machine learning solved complex, specific problems, GenAI’s ability to generate novel outputs, coupled with its cheap, scalable, and versatile nature, has rapidly woven it into the fabric of society. The text emphasizes that for every potential beneficial use, there are equally foreseeable harms, such as the spread of deepfakes and the weaponization of biases. The chapter argues that GenAI is not a neutral tool and that its deployment is often driven by corporate interests that externalize risks onto society. To foster a more informed public discourse, the chapter introduces Neil Postman’s seven questions for evaluating new technology, urging readers to move from being passive subjects to active participants in the democratic oversight of GenAI’s development.

This chapter dissects the intense, seemingly irrational boom in generative AI (GenAI), moving beyond superficial hype to reveal the complex, structural incentives driving billions of dollars into an industry with uncertain returns. It argues that the current boom is propelled not purely by altruistic motives or technological inevitability, but by ten powerful forces. These forces – namely, shareholder primacy, financial imperatives, the pursuit of first-mover advantages, insatiable data-collection needs, self-serving safety justifications, fear of losing relevance, the pursuit of prestige, fear of missing out (FOMO), a belief in “manifest destiny,” and regulatory entrepreneurship – collectively prioritize speed, growth, and market capture over caution, sustainability, and societal benefit. By examining these systemic pressures, the chapter provides the framework for understanding the breakneck pace of AI development and for shaping its ultimate deployment.

The advent of generative AI (GenAI), which can produce billions of words daily, is forcing a profound reexamination of the First Amendment’s protection of “freedom of speech.” Written in 1791 to safeguard human expression, this constitutional right is now being claimed by AI companies seeking immunity from liability for the harmful outputs of their products.

The central question is whether the algorithmic output of a large language model (LLM) – a statistical calculation rather than a human expression of belief or intent – qualifies as “speech” under the First Amendment. Examining the traditional justifications for protected speech – self-expression and the “marketplace of ideas” – the chapter argues that neither rationale neatly applies to content produced by nonsentient machines. It then explores two proposed legal standards: the “speech certainty” standard, which grants protection only when the content is predictable and controllable by a human author, and a human-centric approach requiring intentionality, self-awareness, and human expression. Granting full First Amendment protection could immunize AI companies from liability for a range of harms.

This chapter challenges the notion that generative AI (GenAI) outputs qualify for First Amendment protection as “speech.” It argues that under both speech certainty and human-centric legal frameworks, GenAI outputs – being probabilistic, statistically driven, and devoid of human understanding, sentience, or intent – should be legally categorized as nonexpressive conduct or utilitarian data processing, not intentional communication.

The chapter dismantles the common counterargument that a user has a constitutional “right to receive” an AI’s output, asserting that if no speech is created by the speaker (the AI model), there is no speech to receive. This classification determines the government’s regulatory authority. If AI outputs are deemed speech, regulations face the nearly insurmountable barrier of “strict scrutiny.” If they are nonexpressive conduct, however, the government retains the power to regulate GenAI for safety. The chapter warns that extending constitutional free speech protections to nonhuman, nonintentional algorithms would jeopardize product liability, consumer protections, and democratic self-governance.

This chapter discusses the contentious world of web scraping and its essential, yet problematic, role as the lifeblood of generative AI (GenAI) model training. It highlights the vast scale of data required and the subsequent reliance on the internet’s reservoirs: the public, deep, and dark webs. The central tension explored is the conflict between technological capability and creator consent, revolving around legal and ethical dilemmas.

The chapter scrutinizes the ethical implications of this “finders keepers” approach, the use of torrenting to acquire copyrighted material, and the economic imbalance created when AI models compete with the original content creators. Finally, the chapter examines the complex debate surrounding content opt-out mechanisms and concludes by stressing the impending challenge of diminishing data, which will likely force a greater reliance on synthetic data. Ultimately, the chapter argues that the current approach has created a crisis of legitimacy that requires a reevaluation of the terms under which the internet should be scraped.

This chapter charts the author’s personal and professional journey from an evangelist for AI to an AI critic. It begins by exploring the binary visions of AI’s future: the utopian promise of solving global challenges (curing cancer, reversing global warming) against the existential danger of narrowly optimized superintelligence. The narrative details the author’s tenure at the Allen Institute for Artificial Intelligence, where initial excitement over a commitment to “AI for the common good” and ethical development turned to disillusionment as core ethical initiatives were deprioritized, licenses were changed, and the organization’s focus seemed to shift, and his time as a lecturer on AI, law, and ethics at the University of Texas at Austin. This experience propelled the author into a deep dive into the burgeoning legal and ethical challenges of generative AI (GenAI). The chapter concludes by arguing that GenAI is the decade’s most consequential technology, yet its rapid, legally ambiguous rise is largely outside public consciousness. It makes a forceful call for greater AI literacy and public involvement to ensure that society, rather than a handful of companies, shapes how this technology transforms our world.

The rapid expansion of generative AI (GenAI) is creating an environmental crisis, driven by the massive, opaque infrastructure of data centers. These facilities consume staggering amounts of electricity and fresh water for cooling, accelerating water shortages in drought-prone regions and driving a surge in overall energy demand. This massive demand contradicts claims of AI efficiency, leading to the revival and long-term expansion of fossil fuel infrastructure to support the grid. Furthermore, rising energy costs are being subsidized by ordinary Americans, whose utility bills are increasing because of incentives offered to trillion-dollar AI corporations. A critical problem is the near-total lack of mandatory public reporting on energy and water consumption by the major AI companies, making genuine environmental impact assessment impossible. The chapter concludes that the most important, unanswered question is whether AI’s societal benefits truly outweigh its profound and accelerating costs to the planet and taxpayers.

This chapter examines the collision between generative AI (GenAI) and traditional tort law, specifically with respect to negligence and product liability. It presents the legal dilemma of who is accountable when probabilistic, “black-box” algorithms cause harm. Courts struggle to apply concepts like duty of care and causation to nondeterministic systems. The discussion contrasts negligence, which focuses on a “reasonably prudent developer,” with strict liability, which is limited by whether GenAI is defined as a “product” or a “service.” Key legal frameworks explored include the Learned Hand formula (B < PL) and the “cheapest cost avoider” approach. Ultimately, the chapter argues that courts must evolve past rigid tests, prioritizing functional analysis and expanded developer accountability to deter wrongdoing and prevent catastrophic risks.

This chapter examines how aggressive web scraping by generative AI (GenAI) bots is disrupting the economics of the internet. Unlike traditional indexers, GenAI models indiscriminately ingest massive amounts of data for training (e.g., Meta’s Llama 4 was trained on thirty to sixty trillion tokens). This digital gatekeeping, exemplified by poor crawl-to-referral ratios and the click-reducing impact of Google’s AI overviews, systematically cuts off content creators from the economic benefits of their work. Traditional measures like robots.txt have proven ineffectual. The resulting structural injustice forces websites into an impossible choice – disappear or accept traffic-reducing scraping – threatening the financial viability, quality, and long-term sustainability of the content ecosystem and collective knowledge.

This chapter explores the most significant legal challenge facing generative AI (GenAI) companies: the wave of copyright infringement lawsuits based on unauthorized copying of copyrighted works for model training. It establishes copyright law as the dominant legal battleground, compelling GenAI entities to rely almost entirely on the affirmative defense of fair use.

The analysis reveals two critical flaws in the industry’s fair use claims, challenging the notion of “GenAI exceptionalism”: The Legal Argument is that unlike prior cases of transformative use (e.g., Google Books), GenAI’s core function is shown to exploit the expressive quality of high-quality training material. The Substantive Argument put forth to justify GenAI’s fair use would, if consistently applied, logically grant every human the same right to copy and consume any copyrighted work without payment.

The chapter proceeds to systematically dismantle eight of the most common “substantive arguments” favoring GenAI, concluding that these claims either crumble under scrutiny or serve as thinly veiled demands for an exceptional right to corporate-scale content piracy.

This chapter challenges the popular narrative that open-source software will democratize generative AI (GenAI) and disrupt the existing oligopoly dominated by companies like Google, Microsoft, Meta, and OpenAI. It argues that Big Tech routinely “weaponizes” open-source projects (e.g., Linux, Android, PyTorch) as a strategic tool to gain a competitive advantage over rivals, not for true democratization. The analysis shows that despite impressive percentage growth from startups, the absolute market value gains of tech giants have dwarfed those of new entrants. The enduring concentration of power stems from eight key levers favoring Big Tech: vast capital, exclusive proprietary datasets, massive access to compute (GPUs), network effects, control over top talent, market access agreements, a high-interest investment environment, and a lack of a viable business model to unseat their integrated ecosystems. Ultimately, the chapter concludes that open source in GenAI acts as a “fig leaf” for corporate dominance, as these massive companies control nearly every link in the GenAI supply chain – making radical openness unlikely to have a meaningful impact on the existing power structure.

This chapter critically examines the prevalent practice of measuring generative AI (GenAI) capabilities through performance benchmarks, which major AI companies frequently cite to assert the superiority of their models. It begins by defining benchmarks as standardized tests used to compare models and confer “state-of-the-art” (SOTA) status, despite the tenuous link between high scores and true general ability. The discussion then details the three primary evaluation methods: automated benchmarking, humans as judges, and models as judges.

The chapter then pivots to the numerous problems that undermine the utility of these benchmarks. These limitations include data contamination (training models on the test data), overfitting, inherent dataset and measurement biases, the narrow focus on incrementalism over fundamental research, and the misleading nature of aggregated leaderboards. Ultimately, the chapter argues that benchmarks suffer from a lack of construct validity, testing only what is easily measurable rather than the messy, ill-defined problems of the real world, leading to an over-acceptance of GenAI in critical life domains.

This chapter examines the collision between generative AI (GenAI) model training and foundational contract law. The central claim is that GenAI companies, through their mass data-scraping bots, acquire “actual notice” of website terms of service (ToS) and are therefore legally bound, exposing them to millions of simultaneous breach-of-contract claims. However, content creators’ contract claims, which often prohibit AI training, are frequently preempted by federal copyright law. If forced to pursue a copyright infringement claim, the GenAI company’s fair use defense often prevails. This legal conflict leaves creators without effective recourse, allowing unauthorized use of their work despite clear contractual prohibitions.