# Chapter 6: Intellectual Property, Innovation, and Information Goods
## The Internet as a Copying Technology
Before digital networks, copying and distributing recorded music required physical inputs. A producer pressed records or compact discs, printed packaging, shipped units to stores, and tried to predict how many copies listeners would buy. Unauthorized copying existed, but making and distributing a large number of high-quality copies took equipment, materials, time, and access to customers.
The Internet changed those costs. Once a song had been converted into a digital file, another copy could be made at almost no additional cost. That copy could be sent across town or around the world without a factory, truck, warehouse, or retail shelf. The listener who shared the file did not lose the original. A million people could possess substantially the same recording at the same time.
This was good news if the only objective was to spread music. It was a serious problem if musicians, songwriters, producers, and record companies expected sales of copies to recover the cost of creating, recording, marketing, and discovering music. The first commercially useful recording could be expensive. The next digital copy was nearly free.
It does not follow that every unauthorized download equals one lost sale. Some listeners would have purchased the recording; some would not have purchased at the available price; some may discover an artist and later buy concert tickets or other products. Those differences matter when measuring harm. They do not erase the appropriability problem. A market can be weakened even when the number of copies does not translate one-for-one into displaced purchases.
That contrast is the **first-copy/next-copy problem**. Information can be costly to create and cheap to reproduce. The seller would like a price high enough to recover the first-copy cost. Once the work exists, however, distributing another copy may consume almost no additional resources. A price near marginal cost promotes access but may produce too little revenue to support creation. A higher price can support creation but excludes some users who value access more than the cost of serving them.
Copyright law did not disappear when copying became easier. Unauthorized reproduction and distribution could still violate legal rights. But legal prohibition did not restore the old copying cost. Enforcement against millions of users was expensive, unpopular, incomplete, and sometimes technologically difficult. The change in technology altered the practical meaning of the right.
The recorded-music business adapted. Licensed downloads reconstructed the sale of individual copies in digital form. Streaming went further. A subscriber ordinarily pays for access to a service rather than ownership of a collection of transferable files. Copyright, contracts, subscriptions, authentication, platform organization, and software work together to control access and allocate revenue.
Streaming did not simply return the industry to its earlier state. It changed pricing, distribution, discovery, intermediation, and bargaining. Listeners received broad catalogs and convenient access. Services and rights holders developed new systems for licensing and payment. At the same time, disputes remained about revenue division, bargaining power, artist compensation, and the effects of platform recommendations.
The important lesson is not that the Internet destroyed intellectual property or that streaming solved it. The lesson is that a legal right operates inside an economic and technological environment. When copying, search, monitoring, or distribution costs change, old rights can produce new behavior. Firms and lawmakers then search for a different combination of law, contract, organization, and code.
Recorded music introduces the question that organizes this chapter:
> How should legal institutions govern information that is costly to create, valuable to many people, and easy to copy?
## The Economics of Information Goods
Land, oil, and automobiles are rival resources. If one person consumes a gallon of oil, another person cannot consume the same gallon. If one driver uses an automobile at a particular moment, a second driver ordinarily cannot use it for a different trip at that moment.
Information often behaves differently. One student can learn a formula without preventing another student from learning it. A programmer can use an algorithm while another programmer uses the same idea. A reader who downloads a digital book does not erase the author's copy.
::: quickconcept
**Information Goods**
Information goods include ideas, expression, designs, data, and software that can often be copied or shared at very low marginal cost once they have been created. Creation, verification, marketing, storage, and discovery may still be costly.
:::
Economists call a good **nonrival** when one person's use does not substantially reduce another person's ability to use it. Nonrivalry creates enormous potential gains from sharing. Once a useful idea exists, allowing another person to learn it may create value at very low additional social cost.
Yet easy sharing creates a problem of **appropriability**. A creator may produce benefits that are difficult to convert into income. If competitors can copy a new product immediately, sell it without bearing the development cost, and charge a lower price, the original creator may not recover the investment. Anticipating that outcome, some valuable works and inventions may never be produced.
This gives intellectual property an **incentive-access tradeoff**.
- **Access today:** Existing information should be used widely when the additional cost of access is low.
- **Creation tomorrow:** Creators need some way to recover costs and earn returns when society wants continued investment in new information.
The first objective is sometimes called static efficiency. Given the works and inventions that already exist, broad access can increase current surplus. The second is part of dynamic efficiency. Rules affect which works, inventions, and improvements will exist in the future.
::: keypoint
**The IP Tradeoff**
Intellectual-property rights can encourage creation by helping creators and inventors earn returns. Rights that are too broad, too long, too uncertain, or too costly to license can restrict access, competition, and follow-on innovation.
:::
Neither objective mechanically defeats the other. Free access is not automatically efficient if it causes valuable creation to disappear. Maximum protection is not automatically efficient if it blocks uses that cost almost nothing to permit. The comparison must include realistic alternatives: lead time, secrecy, reputation, contracts, prizes, complementary services, open collaboration, and technological control.
Pricing and organization can soften the tradeoff. A library buys access that many readers share. A subscription spreads a fixed fee across a large catalog and many uses. A firm may charge different prices to students, businesses, and other users when their willingness to pay differs. Advertising or sponsorship can finance access without charging every user directly. These arrangements can expand access while recovering some fixed cost, although each requires information, enforcement, or an intermediary that may possess bargaining power.
Information goods therefore generate more than a copying problem. They create a financing problem: how can society support costly first copies without wasting the low-cost opportunities created by the next copies? Intellectual property is one answer, not the only answer.
## Designing Rights Rather Than Maximizing Them
Debates about intellectual property are often reduced to a single question: Should protection be stronger or weaker? That language is too simple. An IP regime has several dimensions, and strengthening one dimension may produce different effects from strengthening another.
Start with **subject matter**. Which kinds of creation can receive protection? Copyright, patent, trademark, and trade-secret law do not cover the same things because expression, invention, source identification, and confidential information create different problems.
Next consider the **eligibility threshold**. Copyright protection begins from original expression, while a patent requires a more demanding examination of a claimed invention. A low threshold makes protection easier to obtain. It can also create more rights for later users to identify and avoid.
**Breadth** determines how far the right extends. A narrow right may protect one particular implementation. A broad right may reach substitutes, variations, or later improvements. Breadth can increase the expected reward to the first creator while increasing the cost imposed on later creators.
**Duration** determines how long the right lasts. A longer right can increase expected returns, but years far in the future may add little to today's investment incentive while delaying competition and public access.
**Notice** asks whether outsiders can discover the right and understand its boundary. A right that is hard to find or interpret can produce accidental violations, defensive searches, litigation, and abandoned projects.
**Exceptions** preserve activities that would otherwise fall within the right. Fair use, independent creation, and lawful reverse engineering are examples of boundaries that leave room for learning, competition, criticism, or cumulative work.
**Remedies** matter too. An injunction can stop continued use and give a rights holder substantial bargaining leverage. Damages permit use to continue at a legally determined price after infringement is established. Different remedies create different incentives to search, bargain, challenge uncertain rights, and invest in products that combine many components.
Finally, **licensing rules and institutions** determine how permission can be granted. A clear right can make an information asset easier to trade. A fragmented set of uncertain rights can make assembly difficult even when every rights holder would benefit from a completed product.
These dimensions interact. A relatively broad right with a short effective life differs from a narrow right lasting much longer. A broad right with clear notice differs from one whose boundary becomes visible only after litigation. A right paired with standardized licensing differs from one that requires separate negotiations with hundreds of owners.
### A Right Is Not Automatically a Market Monopoly
An intellectual-property right gives control over defined subject matter. It does not necessarily give its owner substantial power over a market. A copyright holder may control one novel while competing with thousands of other books. A patented component may face close technological substitutes. A trademark can distinguish one seller without preventing rivals from selling the same type of product under their own names.
Market power depends on available substitutes, entry, switching costs, demand, complementary assets, and the breadth of the legal right. A narrow patent on a minor feature may have little economic importance. A right covering an indispensable standard, medicine, or bottleneck can provide substantial leverage. The fact that price exceeds the marginal cost of copying also does not by itself prove a harmful monopoly; some markup may be the mechanism used to recover fixed creation cost.
This distinction prevents two errors. One is assuming that every IP right creates a dominant firm. The other is assuming that a formally limited right cannot contribute to market power when alternatives are weak. Chapter 14 examines those competition questions in detail.
The economic objective is therefore not maximum exclusion. It is an institutional package that creates useful incentives while preserving access, competition, and future creativity at reasonable administrative cost.
## Four Different Intellectual-Property Institutions
The phrase **intellectual property** groups together bodies of law with different functions. Treating all of them as ownership of ideas hides more than it reveals.
| Regime | Protects | Economic function | Principal boundary |
| --- | --- | --- | --- |
| Copyright | Original expression | Supports creation and distribution | Ideas, facts, and independent creation remain free |
| Patent | Claimed invention | Rewards invention and disclosure | Limited claims, standards, and term |
| Trademark | Source identifiers | Reduces confusion and search costs | No ownership of the product category |
| Trade secret | Valuable secret information | Rewards secrecy and investment | Independent discovery and lawful reverse engineering |
**Table 6.1. Four intellectual-property institutions.** The regimes govern different subjects and solve different information problems. A single product can involve several regimes at once.
Consider a new smartphone application. Copyright may protect its code, text, images, and music against prohibited copying. Patents may cover qualifying technical inventions used in the system. Trademark protects the name or symbol that tells consumers who supplies it. Trade-secret law may protect valuable confidential methods, data arrangements, or business plans. Contracts govern employees, users, vendors, and licensees. Technical systems determine who can access the service and what they can do with it.
The application is not protected by one abstract right called IP. It sits inside a layered institutional arrangement. Each layer has a different boundary and justification.
## Copyright, Access, and Reuse
Copyright protects original expression, including writings, music, images, films, and software code. It does not give an author ownership of every idea, fact, method, or system communicated by the work.
That distinction is essential. Suppose an author writes a mystery novel set in a remote hotel. Another writer may not copy the novel's text. But copyright does not ordinarily prevent every later writer from using the general idea of a mystery in a remote hotel. If it did, an early author could remove a large field of future stories from the creative commons.
The line between idea and expression can be difficult to draw. A plot can be described at many levels of detail. A software interface can embody creative choices while also allowing users and programs to perform a function. Protect too little and a copier may appropriate most of the value while changing superficial details. Protect too much and the first author may control a genre, method, command structure, or building block that others need to create new works.
Copyright also generally requires copying. Independent creation matters. If two photographers independently capture similar images, similarity alone does not show that one copied the other. In practice, exact independent duplication of a long novel is extraordinarily unlikely. Functional software presents harder settings because different programmers trying to achieve compatibility may produce similar structures or outputs.
This makes copyright boundaries comparatively sharp near the center and less certain near the edge. Reproducing an entire book without permission is easy to classify. Disputes over characters, interfaces, structure, or nonliteral similarity demand more interpretation. Friedman's useful intuition is that copyright resembles a clear fence when it targets literal copying, but becomes less clear as protection moves toward function or general design.
### Fair Use as Breathing Room
Copyright cannot serve creation by preventing criticism, teaching, research, parody, quotation, and every reuse that incorporates protected material. U.S. law therefore recognizes **fair use**, a context-specific inquiry that considers the purpose and character of the use, the nature of the copyrighted work, the amount and importance used, and the effect on relevant markets.
::: sideline
**Fair Use Is Not A Percentage Rule**
Educational purpose can matter, but it does not make every classroom use fair. Nor is there a fixed number of words, seconds, pages, or percentage that is always permissible. Fair use requires a case-specific balance of several considerations.
:::
Economically, fair use can reduce transaction costs and preserve room for uses whose social value is high relative to the threat they pose to creation incentives. A book reviewer may need to quote a passage to criticize it. A historian may need to reproduce part of a document to explain it. Requiring a negotiated license for every quotation could let a rights holder suppress criticism or make many small uses too costly to clear.
The flexibility of fair use is also a cost. A bright rule would be easier to predict, but it could perform poorly across books, music, software, search, parody, teaching, and technologies not yet invented. A flexible standard adapts, but parties may disagree about its application and litigate the boundary.
The **public domain** supplies another form of breathing room. Works and elements not protected by copyright can be used without permission. Public-domain material is not wasted material. It is an input into new editions, performances, adaptations, archives, databases, and creative works.
Copyright also creates **tracing costs**. A potential user may need to determine whether a work is protected, who owns the relevant rights, whether several rights are involved, and whether the intended use requires permission. The search may be easy for a current commercial book and difficult for an older photograph whose creator or chain of ownership is unclear. A right that cannot be reliably traced can discourage valuable use without delivering a corresponding payment to the owner.
Registries, collective licensing, standardized permissions, and reliable ownership records can reduce tracing costs. Their quality is part of copyright policy even though they do not change the nominal scope of the right.
### Law, Contract, and Code
Digital works are often governed by more than copyright. A streaming service uses contracts to define subscriber access, software to authenticate users, and code to limit copying or redistribution. A publisher may combine copyright with encryption or digital-rights management.
Technological control can make exclusion cheaper and more immediate. It can also govern conduct more strictly than copyright alone. A technological system may prevent a use that would have been lawful, make repair or interoperability harder, or eliminate practical access when a service closes. Contract and code can therefore fill enforcement gaps while creating a new question: Should privately designed controls determine every use they can technically prevent?
That question returns later in the chapters on platforms and smart contracts. For now, note the institutional choice. Copyright supplies public legal rules. Contracts tailor permissions among parties. Code enforces some permissions automatically. Each tool changes enforcement cost, error, flexibility, and user bargaining power.
## Patents, Boundaries, and Cumulative Innovation
A patent concerns a claimed invention rather than a particular physical object. The owner receives a time-limited right to exclude others from specified conduct involving the claimed invention. A patent is not an affirmative government license to make or sell a product. Another patent, safety regulation, or other law may still prevent the owner from practicing the invention.
::: historicalnote
**The Patent Bargain**
Patent law is often described as a bargain. An inventor discloses a qualifying invention, and the public grants a temporary right to exclude. When the right ends, the disclosed invention remains available for public use, subject to other valid rights and rules.
:::
The bargain has several possible economic benefits. Expected exclusivity can support costly research and development. Disclosure can spread technical knowledge. Publication of a claim can help others identify the invention, license it, invent around it, or build on it after the right expires.
But patents create boundaries in what Friedman calls **idea space**. An inventor does not merely point to a machine and say, "I own this object." Patent claims describe the territory the inventor asserts. A narrow claim may be easy for competitors to avoid. A broad claim may cover substitutes and future implementations the inventor did not build.
Patent examination attempts to reject claims that fail requirements such as novelty and nonobviousness. The process improves notice and screens applications, but it does not eliminate uncertainty. Technical language can be difficult. Relevant earlier work may be missed. Courts may later interpret a term differently, identify invalidating prior art, or disagree about whether a product falls inside the claim.
### Breadth, Duration, and the Amazon Illustration
Suppose an inventor develops a particular way to reduce an online purchase to one customer action. Several possible boundaries can be imagined:
1. protection for the inventor's precise implementation
2. protection for a defined class of one-action ordering systems
3. protection for the general objective of completing an online purchase with one action
The first boundary may be so narrow that a rival can imitate the economic function with a minor technical change. The third may be so broad that it controls a general commercial objective and blocks many later designs. The middle boundary may also be difficult to define.
Amazon received a U.S. patent associated with one-click ordering and used it in litigation against Barnes & Noble. In an appeal involving a preliminary injunction, the court concluded that Barnes & Noble had raised a substantial question about validity. The court did not finally hold the patent invalid in that decision.
The episode matters here as a design problem, not as a detailed case history. How much territory in idea space should an early online retailer receive? A patent too narrow may provide little reward. A patent too broad may transfer control over a general goal rather than a specific contribution.
Duration creates a related tradeoff. A longer term increases expected returns, especially when a valuable invention remains commercially useful. It also delays unrestricted competition. Economic models show that different combinations of breadth and duration can create similar expected rewards while producing different losses from exclusion. There is no general reason to assume that increasing both dimensions always improves innovation.
### Patent Races and Follow-On Invention
Imagine two research teams approaching the same discovery. Team A finishes in January. Team B would have finished in July even if Team A had never existed. A long patent awarded to Team A may create a large private return even though the patent accelerated the invention by only six months.
The expected prize can motivate useful effort. It can also cause several teams to duplicate research in a race to finish first. Some duplication is productive because competing approaches generate information and increase the probability of success. Too much race expenditure can consume most of the social value created by arriving slightly earlier.
Innovation is also cumulative. A new drug may build on earlier biological discoveries, research tools, and compounds. A software product may combine protocols, compression methods, interfaces, security techniques, and thousands of code modules. An improvement can be patentable while still depending on permission to use an earlier patented invention.
When many complementary rights are fragmented among owners, a follow-on innovator must search for them, determine which are valid and relevant, negotiate licenses, and manage the risk that another claim appears after investment. This is the **patent-thicket** problem. The problem is not merely that many patents exist. It arises when overlapping, uncertain, or complementary rights make assembly expensive.
Patent pools, standardized licenses, clearinghouses, cross-licenses, and carefully chosen remedies can lower assembly costs. They can also create opportunities for collusion, exclusion, or strategic portfolio building. Once again, the institutional comparison matters more than the label.
### Notice, Remedies, and Bargaining Leverage
Patent notice is difficult because inventors may independently arrive at a technology without reading a competitor's patent. A product with thousands of components may implicate claims from many technical fields. Searching every possible claim can be costly, yet failing to search can leave a firm exposed after it has invested in design, factories, customers, and compatibility.
Remedies affect bargaining in that setting. If an injunction can stop an entire product because one small component infringes, the rights holder may bargain over the value of the completed product rather than the value contributed by the component. Damages can reduce that holdout leverage, but courts must then estimate a payment and preserve adequate incentives to license before using the invention.
The timing of negotiation matters. A license discussed before product design may reflect available substitutes. A demand made after the product is locked in can capture switching costs. This does not mean that late-asserted rights are invalid or that injunctions are always inappropriate. It means that patent remedies shape transaction costs, investment incentives, and bargaining baselines, not merely compensation after a wrong.
### Pharmaceuticals and Software
Uniform patent rules can operate differently across innovation environments. The following comparison is deliberately stylized. Individual products and firms vary within both industries.
| Feature | Pharmaceuticals | Software |
| --- | --- | --- |
| Development path | Long testing and regulatory review | Shorter, iterative releases |
| Product cycle | Often long | Often rapid |
| Invention boundary | Molecules and uses can be discrete | Functions may overlap and combine |
| Cumulative dependence | Important, often sequential | Dense and highly modular |
| Other returns | Regulation, know-how, marketing | Lead time, services, data, networks |
| Main patent risk | Access and delayed competition | Thickets and blocked entry |
**Table 6.2. Two innovation environments.** The comparison explains why the same nominal patent rules can create different incentive, notice, and assembly effects. It does not classify all pharmaceutical patents as desirable or all software patents as harmful.
A pharmaceutical developer may face long research, testing, and regulatory processes before earning revenue. Once a compound and its use are known, some forms of imitation may be easier than the original development path. Patents can therefore be important to expected returns. The same exclusivity can raise prices and delay competition for products that matter greatly to patients.
Software often develops through shorter cycles, modular combinations, and repeated updates. Lead time, subscriptions, service, data, reputation, compatibility, and network effects may help firms earn returns. A product can also implicate many technical claims, making notice and rights assembly difficult.
These are tendencies, not verdicts. Software can require enormous investment. Pharmaceutical innovation can be densely cumulative, and portfolios can create thickets. The point is that legal rules interact with product cycles, regulation, technical boundaries, complementary assets, and the structure of follow-on innovation. Patent policy is too important to be analyzed as if every industry were the same.
## Trademark and Trade Secret
Trademark law has a different center of gravity from copyright and patent. A trademark identifies the source of goods or services. Names, symbols, and other indicators help consumers connect a product with prior experience, advertising, reputation, and expected quality.
Suppose two restaurants use nearly identical names and signs. Customers may enter the second restaurant believing it is associated with the first. The second restaurant can free ride on the first restaurant's reputation, while poor quality at the second can damage that reputation. Preventing confusing uses can reduce consumer search costs and support investment in quality.
This rationale does not require ownership of every word or product feature. The central question concerns source identification and likely confusion in context. A business ordinarily cannot use trademark to own the general category of the product. If the ordinary name of a product were controlled by one seller, rivals would struggle to tell consumers what they sell.
Trademark can promote competition by making sellers easier to identify. It can also be asserted too broadly. A rule that suppresses nonconfusing comparison, criticism, parody, or ordinary descriptive language would increase search costs rather than reduce them. The economic boundary follows the information function.
Trade-secret law protects valuable information that derives value from not being generally known and is subject to reasonable efforts to maintain secrecy. A formula, manufacturing method, algorithm, customer strategy, or business process may qualify when the legal requirements are met.
Trade secret is partial protection. It targets improper acquisition, disclosure, or use. It does not ordinarily prevent another person from developing the same information independently. Lawful reverse engineering can also remain available, although contracts and other law can complicate particular situations.
This boundary creates a distinctive compromise. A firm receives protection against theft, espionage, bribery, and breach of duties of confidence. Competitors remain free to discover the information through their own research or through lawful examination of a product. The law rewards both secrecy investment and independent innovation.
Trade-secret protection has costs. Secrecy can restrict communication among employees and researchers. Firms may duplicate work because knowledge is not disclosed. Employees may face disputes when general skill and experience overlap with a former employer's confidential information. Unlike a published patent, a secret may provide little public technical disclosure.
The choice between patent and secrecy therefore changes more than duration. A patent requires disclosure and provides a claim against independent invention within its scope. Trade secret preserves secrecy but can disappear when the information becomes lawfully known. A firm chooses among these tools by considering detectability, product life, disclosure, reverse engineering, employee mobility, and the likelihood of independent discovery.
## Licensing and Other Ways to Earn Returns
An intellectual-property right can support innovation without the owner producing every final product. Through a **license**, the rights holder gives another party permission to use protected material under agreed conditions. The license may specify field of use, territory, duration, quality control, payment, access to updates, or other terms.
Licensing can move information toward higher-valued uses. A scientist may invent a process but lack manufacturing capacity. A publisher may know how to edit, market, and distribute an author's work. A small software developer may license technology to firms with established customers. Royalties can divide returns while specialization combines complementary capabilities.
Licensing also consumes resources. Parties must identify each other, evaluate the right, negotiate terms, monitor performance, and resolve disputes. Uncertain scope makes bargaining harder. Fragmented rights create holdouts. A license may restrict competition or interoperability as well as enable production.
Three tasks are especially important. First, the user must find the relevant owners. Second, both sides must determine what the rights cover and how valuable they are. Third, they must agree on price and conditions before uncertainty or strategic delay destroys the cooperative surplus. A market can fail at any of these steps even when permission would make both sides better off.
Collective institutions can reduce repeated negotiation. A clearinghouse can identify rights and process payments. A patent pool can assemble complementary technologies into one package. A standard license can reduce drafting and interpretation costs. These institutions trade customization for scale. They work best when participation, pricing, scope, and governance prevent the coordinating body from becoming a new bottleneck.
Formal IP is only one way to earn returns from information.
::: quickconcept
**Alternatives To Formal IP**
Creators and firms may rely on lead time, secrecy, reputation, contracts, complementary services, technological controls, prizes, or open collaboration instead of, or alongside, formal intellectual-property rights.
:::
**Lead time** lets an innovator sell before competitors catch up. **Reputation** helps creators and firms attract customers to later work. **Complementary assets** such as manufacturing, service, distribution, data, hardware, and customer relationships can remain scarce even when an idea is easy to copy. **Technological controls** can make access conditional. **Contracts** can restrict disclosure or use among parties.
Governments, foundations, or buyers can also use **prizes**. A prize specifies a goal and rewards a successful solution. Compared with a patent, a prize can make the solution broadly available after payment. The difficulty is setting the reward and defining success before the solution and its value are known. Prizes and patents can therefore complement rather than simply replace one another.
## Open Source: Innovation Without Maximum Exclusion
Open-source software demonstrates that the alternative to maximum exclusion is not necessarily no governance. Copyright gives software creators legal authority over copying, modification, and distribution. An open-source license uses that authority to grant broad permissions under stated conditions.
::: keypoint
**Open Source Uses Intellectual Property**
Open-source licenses use copyright permissions and conditions to organize access, modification, and redistribution. Openness is deliberately structured rather than legally empty.
:::
Three broad models help organize the comparison.
| Model | User permissions | Redistribution condition | Common revenue path |
| --- | --- | --- | --- |
| Proprietary | Limited by license | Source and modification usually restricted | License, subscription, complements |
| Permissive open source | Broad use and modification | Notices and license terms retained | Services, hosting, integration |
| Copyleft open source | Broad use and modification | Derivatives shared under reciprocal terms | Services, support, complementary products |
**Table 6.3. Software-governance models.** These are broad categories, not individual licenses. Each model combines permissions, conditions, governance, and possible revenue sources differently.
Under a proprietary model, the producer usually limits access to source code and reserves most rights to modify or redistribute the program. Users receive specified permissions through a license or subscription.
A permissive open-source model grants broad authority to inspect, use, modify, and redistribute code while imposing relatively limited continuing conditions. A recipient may often combine the code with a proprietary product, subject to the actual license.
Copyleft also grants broad permissions but requires specified derivative distributions to remain under reciprocal terms. The objective is not merely to release one copy of source code. It is to preserve openness as the code is modified and redistributed.
Linux provides the central example. It is not a software commons with no boundaries or decision-makers. Contributors submit changes through organized processes. Maintainers review code. Projects control repositories, releases, names, and technical direction. Licenses govern redistribution. Firms and volunteers contribute for different reasons, including learning, reputation, shared infrastructure, product compatibility, service revenue, and the need to improve a tool they use.
The legal permission to modify code does not guarantee that a modification enters the main project. Maintainers still decide which changes meet technical, security, and organizational standards. A dissatisfied group may be able to copy the existing code and create a separate version, but doing so can divide contributors, users, documentation, and compatibility. The possibility of a fork constrains governance while the cost of fragmentation gives participants a reason to remain coordinated.
Open source can reduce licensing and entry costs. Many users can inspect the code, identify errors, adapt it to new settings, and build interoperable products. Improvements can accumulate across organizations. Firms need not recreate every shared component from the beginning.
Openness does not solve every incentive problem. Popular projects require maintenance, security review, documentation, coordination, and decisions about which changes to accept. Contributors may free ride on others' work. Projects can fragment into incompatible versions. License terms can conflict. A technically open project may still be governed by a small group controlling the main repository or official release.
Firms can earn returns around open source. They may sell hosting, support, customization, certification, hardware, security, or enterprise integration. They may contribute to shared infrastructure because doing so reduces their own costs or expands demand for complementary products. The business model shifts from charging for every copy toward charging for scarce complements.
This arrangement also changes who finances maintenance. A widely used component may support valuable commercial products while depending on a small group of maintainers. Because every user can hope that someone else will contribute, shared infrastructure can be underfunded. Foundations, corporate sponsorship, paid maintainers, procurement rules, and reputation incentives are institutional responses to that public-goods problem.
Creative Commons offers a parallel for expressive works. A copyright holder can use a standardized license to authorize sharing or reuse under selected conditions. This is not the dominant licensing system for software, and the available conditions differ. The common idea is that an owner can choose a structured permission regime rather than either reserving every right or abandoning control entirely.
Compare open source with the recorded-music opening. Streaming responded to cheap copying by reconstructing controlled access through platforms, subscriptions, licensing, and code. Open source permits copying and governs modification and redistribution. Both are institutional adaptations to low-cost reproduction, but they organize scarcity and openness differently.
## AI and the Next Boundary Problem
Generative AI makes the chapter's older questions newly visible. Models can be trained using large collections of text, images, music, and code. They can then produce outputs that resemble categories of human work, perform related functions, and compete for attention or income.
The debate becomes confused when several questions are collapsed into one.
1. Was protected material copied or otherwise used in a legally relevant way during collection and training?
2. Does a particular output reproduce protected expression, or does it share ideas, facts, functions, techniques, or stylistic features that the law treats differently?
3. Does the output substitute for the original work, complement it, or change demand for future human creation?
4. Would licensing, collective management, technical controls, transparency, or another rule improve incentives and access after counting transaction and enforcement costs?
The first two are legal classification questions informed by facts about the technology. The third is an economic question. The fourth is an institutional-design question. They affect one another, but they are not interchangeable.
::: casestudy
**News Media, Copyright, And AI Training**
News organizations spend resources gathering facts, interviewing sources, verifying claims, editing stories, and maintaining reporting institutions. Published articles can then be copied cheaply and may be valuable as AI training inputs. AI developers argue that training can produce new capabilities rather than substitute copies. Publishers argue that unlicensed use can appropriate valuable expression and support products that compete with their reporting.
Current U.S. disputes over training, fair use, output similarity, and licensing remain unsettled and context-specific. The economic questions are also open. Licensing could compensate creators and support production. It could also require costly rights clearance at scale, advantage large firms, or exclude socially valuable research. A rule favoring unlicensed training could accelerate model development while weakening some creation incentives or bargaining positions. The relevant comparison is among feasible institutions, not between innovation and copyright as abstract absolutes.
:::
AI can sometimes learn what might loosely be called a **reduced form** of a body of work: features that help generate economically substitutable output without reproducing one obvious passage, image, or file. This is an economic metaphor, not a category in copyright law. It highlights a possible gap between conventional copying analysis and market effect.
Economic substitution alone does not prove infringement. Copyright law has its own requirements and boundaries. But absence of literal copying does not prove absence of economic effect. A model can change demand, creation incentives, or bargaining even when an output is not a recognizable copy of one work.
::: sideline
**Why Intellectual Property Needs Economists**
Technology changes copying, search, licensing, exclusion, and invention costs faster than legal categories change. Economic analysis compares how alternative boundaries affect innovation, access, competition, and transaction costs without assuming that stronger or weaker protection is always better.
:::
The AI problem is therefore a culmination of the chapter, not an exception to it. Identify the information good. Identify the right and its boundary. Separate legal status from economic effect. Compare formal exclusivity with licensing, contracts, code, collective institutions, and structured openness. Then ask which errors each arrangement is likely to make.
## Big Picture
Intellectual property is a set of coordination technologies for information goods. It creates legal control where copying may otherwise be easy, but it does not turn ideas and expression into ordinary physical property.
The economic problem begins with the first-copy/next-copy distinction. Broad access to an existing work can be valuable because another user may impose little additional cost. Continued creation can also require a way to recover investment. Good institutional design must consider both margins.
Copyright, patent, trademark, and trade secret respond differently because they solve different problems. Copyright protects expression while preserving room for ideas, facts, independent creation, and fair use. Patent protects claimed inventions but creates difficult questions of scope, duration, notice, and cumulative innovation. Trademark lowers consumer search and confusion. Trade secret protects against improper acquisition while leaving room for independent discovery and lawful reverse engineering.
Rights are only part of the system. Licensing, lead time, secrecy, reputation, prizes, complements, technological controls, open source, and Creative Commons organize returns and access in other ways. Streaming and open source show two different adaptations to cheap copying: one reconstructs controlled access, while the other uses licensing to structure openness.
The central question is not whether intellectual property is good or bad. It is which combination of boundaries, duration, notice, exceptions, remedies, licensing, organization, and technology best supports valuable creation and use under realistic institutional constraints.
## Chapter Study Map
- **Core ideas:** information goods, nonrivalry, first-copy/next-copy cost, appropriability, incentive versus access, static and dynamic efficiency, and multidimensional rights design.
- **Legal institutions:** copyright, patent, trademark, trade secret, licensing, fair use, independent creation, reverse engineering, and technological control.
- **Tables:** distinguish the four IP regimes, compare pharmaceutical and software innovation environments, and compare proprietary, permissive, and copyleft software governance.
- **Reasoning tasks:** identify the information problem, locate the legal boundary, diagnose underprotection or overprotection, and compare formal rights with realistic alternatives.
- **Common mistakes:** treating all IP as one form of ownership, assuming broader rights always increase innovation, confusing economic harm with legal infringement, or treating open source as the absence of rules.
- **Practice examples:** recorded music and streaming, Amazon's one-click ordering process, patent races, cumulative innovation, trademarks, reverse engineering, Linux, Creative Commons, news content, and AI training.
- **Optional enrichment:** detailed fair-use cases, patent-eligibility doctrine, international IP, pharmaceutical patent-term rules, standards-essential patents, and current AI litigation.
## Review Questions
1. What is the first-copy/next-copy problem?
2. How did the Internet change the practical cost of copying and distributing recorded music?
3. Why is streaming an institutional adaptation rather than simply stronger copyright enforcement?
4. What does it mean for an information good to be nonrival?
5. Explain the appropriability problem.
6. Distinguish static access from dynamic creation incentives.
7. Why is stronger versus weaker protection an incomplete way to describe IP design?
8. Name the principal dimensions of an intellectual-property regime.
9. Why does notice matter to the economic cost of a right?
10. How do copyright, patent, trademark, and trade secret address different information problems?
11. Distinguish copyright protection of expression from ownership of an idea.
12. Why does independent creation matter in copyright?
13. Why can fair use reduce transaction costs?
14. Why does fair use also create uncertainty?
15. How can contract and code govern digital works beyond copyright's direct rules?
16. What does a patent owner's right to exclude mean?
17. Why is a patent not an affirmative right to practice an invention?
18. Explain the tradeoff between patent breadth and inventing around.
19. What does the Amazon one-click example teach about claim scope?
20. How can a patent race produce both useful competition and wasteful duplication?
21. What makes a collection of patents a thicket rather than merely a large portfolio?
22. Why can the same patent rules have different effects in pharmaceuticals and software?
23. How can trademark protection reduce consumer search costs?
24. Why is trade-secret protection described as partial protection?
25. How do lead time, reputation, and complementary services help creators earn returns without relying entirely on formal IP?
26. How can a prize differ from a patent as an innovation incentive?
27. Why is open source not the absence of intellectual property?
28. Distinguish permissive open source from copyleft at a principles level.
29. What governance problems remain in an open-source project?
30. Why must legal copying, economic substitution, and institutional design remain separate questions in AI disputes?
## Economic Reasoning Questions
1. A digital textbook costs `\$500,000` to research and produce, but another download costs `\$0.10` to serve. Explain why marginal-cost pricing promotes access but may not finance the first copy. Identify at least three possible institutions that could support production.
2. A musician can earn revenue from recordings, performances, merchandise, patron support, and licensing. How does the availability of these complements affect the case for stronger protection of recordings? What information would you need before recommending a rule?
3. A software interface can be copied cheaply, but compatibility allows users to switch products without retraining. Explain the incentive and competition effects of protecting the interface broadly.
4. Two patent designs generate the same expected reward. Design A is narrow and long; Design B is broad and short. Explain why they may produce different effects on substitutes, follow-on invention, and licensing.
5. An online retailer develops a particular one-action checkout method. Compare protection of the precise code, the implemented system, a class of one-action systems, and the general objective of quick checkout. Which boundary problems arise at each level?
6. Four firms are racing toward the same invention. A patent prize increases their research effort and advances expected discovery by three months. Explain why the private reward may exceed the social value of the acceleration. Does that establish that the race is wasteful?
7. A new device requires licenses to fifteen complementary patents owned by different firms. Explain how search, uncertainty, holdout, and royalty stacking can prevent a project with positive cooperative surplus. Compare one possible assembly institution.
8. A rival legally reverse engineers a product and discovers a production method that the original firm kept secret. Explain why permitting the rival can reduce the original firm's return while still serving an economic function.
9. A company releases a useful software library under a permissive open-source license and earns revenue from hosting and enterprise support. Explain how open copying can increase demand for scarce complements.
10. A copyleft requirement causes improvements to remain open but discourages one firm from combining the code with a proprietary system. Identify the cumulative-innovation benefit and the possible integration cost.
11. An AI system produces summaries that reduce visits to news websites but does not reproduce recognizable passages in its answers. Separate the legal, economic, and institutional-design questions.
12. A proposed AI-training license would compensate publishers but require negotiations with thousands of rights holders. Compare the creation-incentive benefit with transaction-cost, entry, and research effects. What institutions might lower those costs?
## Law and Economics Lab
### Design and Audit an Intellectual-Property Regime
Choose one emerging information technology. Possibilities include an AI research tool, a synthetic-media platform, a medical discovery system, a collaborative software project, a sports-data product, or another technology approved by your instructor.
Your task is to design an institutional regime and then audit it from the perspective of creators, users, follow-on innovators, competitors, and administrators.
1. **Define the information good.** What is costly to create? What can be copied or shared cheaply? Which inputs and complements remain scarce?
2. **State the market failure.** Is the central concern weak appropriability, consumer confusion, secrecy, fragmented rights, costly licensing, or something else?
3. **Design the legal right.** Specify subject matter, eligibility threshold, breadth, duration, notice, exceptions, remedies, and transferability.
4. **Compare alternatives.** Compare your proposed right with at least two of the following: lead time, secrecy, contract, prize, technological control, open source, collective licensing, or complementary-service revenue.
5. **Test cumulative innovation.** Explain how a later creator would discover the boundary, obtain permission when needed, and challenge a mistaken or overbroad claim.
6. **Identify likely errors.** Describe one underprotection error and one overprotection error your regime might produce.
7. **Evaluate distribution.** Identify who receives bargaining power, revenue, access, and risk. Keep this analysis separate from the efficiency comparison.
8. **Audit an AI proposal.** Ask an AI system to recommend a regime. Identify at least three hidden assumptions, verify every legal claim using authoritative sources, and revise the proposal.
Conclude with a short recommendation. Do not argue that your design is perfect. Explain why its expected coordination benefits exceed its information, enforcement, licensing, exclusion, and error costs relative to the best feasible alternative.