Computer Associates International (CA) developed a job scheduling program that relied on an internal interface module called ADAPTER to communicate with different operating systems. A programmer who left CA later joined Altai, Inc. At Altai, he used CA's confidential ADAPTER source code without authorization to help write Altai's analogous interface component, called OSCAR, version 3.4. The district court found that a significant portion of OSCAR 3.4 (roughly one-third) was copied from ADAPTER. After CA discovered the copying and filed suit, Altai undertook a clean-room rewrite and produced OSCAR version 3.5 using programmers who had not been exposed to CA's code. CA nonetheless alleged that OSCAR 3.5 infringed the non-literal elements (structure, sequence, organization, and design choices) of ADAPTER. CA also asserted state-law claims, including trade secret misappropriation and unfair competition. The district court held that OSCAR 3.4 literally infringed; applying an abstraction–filtration–comparison analysis, it concluded that OSCAR 3.5 did not infringe. It further held that certain state-law claims were preempted by the Copyright Act, but that a properly framed trade secret claim premised on breach of confidence/improper means would not be.
How should courts determine whether a computer program infringes the non-literal elements of another, and to what extent are state-law claims such as trade secret misappropriation preempted by the Copyright Act when they arise from the copying of computer programs?
Non-literal infringement of computer software is assessed using the abstraction–filtration–comparison test: (1) Abstract the program into its constituent structural levels; (2) Filter out unprotectable elements, including ideas and processes (17 U.S.C. § 102(b)), elements dictated by efficiency (merger), external constraints (e.g., operating system requirements, hardware standards, industry practices, and compatibility mandates), materials from the public domain, and scènes à faire arising from standard programming practices; and (3) Compare the remaining protectable expression with the accused work for substantial similarity. Regarding preemption, 17 U.S.C. § 301 preempts state-law claims that assert rights equivalent to those protected by copyright; however, claims containing an extra element —such as breach of confidence or improper means in trade secret law—are not preempted because they are qualitatively different from a claim of mere copying.
The Second Circuit adopted the abstraction–filtration–comparison test and affirmed that OSCAR 3.4 infringed CA's copyrighted code through literal copying. It also affirmed that, after proper filtration, OSCAR 3.5 did not infringe CA's non-literal program elements. On preemption, the court held that state-law claims equivalent to copyright rights are preempted, but a trade secret misappropriation claim predicated on breach of confidence or improper means is not.
The court recognized that software differs from traditional literary works because much of its content is functional and constrained by technical requirements. A simplistic inquiry into overall structure, sequence, and organization risks affording copyright protection to ideas, processes, and methods of operation barred by § 102(b). Critiquing broader approaches like Whelan v. Jaslow, the court endorsed a more granular methodology that disaggregates a program into hierarchical levels—from high-level architecture down to modules and source code—so courts can isolate expression from function. In the filtration step, the court excluded elements dictated by efficiency (merger), externalities (e.g., operating system interfaces, industry standards, hardware constraints, and compatibility requirements), standard programming practices (scènes à faire), and public-domain material. Only after removing these did the court compare what remained of ADAPTER's protectable expression against OSCAR 3.5. That comparison revealed insufficient similarity to establish infringement, particularly given Altai's clean-room development and the predominance of functional and externally constrained elements in the overlapping features. On preemption, the court applied § 301's equivalency test: state-law claims that add no element beyond acts within the scope of exclusive rights under § 106 are preempted. Unfair competition claims premised solely on copying were therefore displaced. By contrast, trade secret misappropriation, which requires proof of breach of confidence or other improper means, demands an extra element and thus survives preemption. The court's approach harmonizes federal copyright policy—ensuring robust protection of original expression while leaving ideas, processes, and technological interfaces free for competition—with state protection against wrongful acquisition or disclosure of confidential know-how.
Altai is the leading case on non-literal software copyright infringement. Its abstraction–filtration–comparison test has been widely adopted because it operationalizes the idea/expression dichotomy for software and carefully cabins copyright to expressive choices rather than functional necessities. The decision also provides a clear framework for preemption analysis, distinguishing between preempted state claims that mirror copyright and non-preempted trade secret claims involving breach of confidence or other improper means. For law students, Altai is essential to understanding how courts balance innovation incentives with interoperability, efficiency, and competition in the software industry.
Computer Associates v. Altai established a disciplined, technology-aware method for adjudicating software copyright disputes. By requiring courts to abstract, filter, and then compare, the decision ensures that copyright remains tethered to expressive authorship and does not extend to ideas, functionality, or constraints necessary for efficient and interoperable computing.