Artificial intelligence is rapidly changing not only how Michigan businesses perform work, but also what they regard as valuable proprietary information. A company may spend months refining prompts that consistently produce useful results from a generative AI system. A law firm may develop a library of structured prompts for analyzing contracts, preparing deposition outlines, summarizing records, or identifying issues in discovery. A manufacturer may create an AI-assisted workflow that combines proprietary production data, carefully designed instructions, validation rules, human review, and specialized software integrations. A marketing company may develop prompting sequences that turn internal customer information into highly targeted campaigns. As these practices mature, a practical legal question follows: can the prompts, prompt libraries, and workflows themselves be protected as trade secrets?
Under Michigan law, the answer can be yes. A prompt does not become a trade secret merely because someone labels it confidential, however, and an AI workflow does not receive legal protection simply because substantial time was spent creating it. Protection depends on the requirements imposed by the Michigan Uniform Trade Secrets Act, commonly referred to as MUTSA. Michigan’s statute defines a trade secret broadly to include information such as a formula, pattern, compilation, program, device, method, technique, or process when the information derives independent economic value from not being generally known or readily ascertainable by proper means and is subject to reasonable efforts to maintain its secrecy. ¹ Those categories are broad enough to encompass many of the information structures businesses now create around generative AI.
There does not appear to be a published Michigan appellate decision squarely deciding whether a generative-AI prompt, a proprietary prompt library, or a particular AI workflow is a trade secret. That does not mean such assets fall outside existing law. Trade-secret statutes were deliberately written to protect valuable confidential information without limiting protection to a particular technological format. The federal Defend Trade Secrets Act likewise defines trade secrets to include compilations, methods, techniques, processes, procedures, programs, and codes, whether tangible or intangible and regardless of how they are stored. ² The absence of a Michigan case specifically using the term “prompt engineering” therefore should not be confused with an absence of legal protection. Courts can apply established trade-secret principles to new forms of commercially valuable information.
One of the most important starting points is that trade-secret law generally does not depend on whether information qualifies for patent or copyright protection. A company does not have to establish that a prompt is a patentable invention or a copyrightable literary work before arguing that the prompt is a trade secret. The inquiry is different. Trade-secret law asks whether the information has economic value because it is not generally known and whether reasonable steps have been taken to maintain its secrecy. ¹ A relatively short body of text may therefore be commercially significant if it encodes knowledge that competitors do not possess and cannot readily reproduce.
This distinction matters because individual prompts can appear deceptively simple. A finished prompt might consist of only several paragraphs of instructions. Viewed in isolation, some of its language may be ordinary. The commercial value, however, may lie in the selection of instructions, the sequence in which they are presented, the examples embedded in the prompt, the variables passed into it, the criteria used to evaluate the output, the model settings applied to it, or the way the prompt is integrated with proprietary business information. A trade secret can reside in the architecture of a solution rather than in an individual sentence.
Michigan and Sixth Circuit authorities support this broader way of analyzing proprietary information. In Mike’s Train House, Inc. v. Lionel, LLC, the Sixth Circuit, applying Michigan trade-secret law, considered design drawings containing both secret and nonsecret information. The court recognized that a trade secret may exist in a unique combination even where individual elements of that combination are publicly known.⁴ That principle is particularly important for AI prompting. A company may use common words, conventional instructions, publicly known prompting techniques, and commercially available AI models, yet still develop a proprietary combination that produces results competitors cannot easily replicate.
The Seventh Circuit’s decision in 3M v. Pribyl, which the Sixth Circuit discussed approvingly in Mike’s Train House, illustrates the same concept. There, the court explained that a protectable secret can arise from the unified combination and operation of characteristics or components even when individual portions are in the public domain.⁹ Applied to AI, the potential trade secret may therefore be the integrated prompting system rather than a supposedly magical phrase hidden inside one prompt.
An individual prompt has the strongest claim to trade-secret protection when it represents more than a generic request to an AI system. A prompt such as “draft a professional letter” is ordinarily too commonplace to possess meaningful secrecy-based economic value. A highly refined instruction set designed through extensive testing may present a different case. For example, a company may discover that a particular structure of instructions, examples, exclusions, definitions, reasoning constraints, formatting requirements, and quality-control steps produces substantially more reliable output for an important recurring business task. If competitors do not know that structure and cannot readily discover it through legitimate means, the economic-value requirement becomes easier to establish.
The relevant value need not come from the number of words in the prompt. It may come from the time the prompt saves, the errors it prevents, the revenue it helps generate, the expertise it captures, or the consistency it creates. A prompt that reduces a three-hour task to twenty minutes can be economically significant even if it fits on a single page. Likewise, a prompt that incorporates years of institutional knowledge into repeatable instructions may have commercial value far exceeding the apparent complexity of the text itself.
Evidence of development can become important. A business that can show repeated testing, rejected versions, performance comparisons, internal benchmarks, specialist input, or measurable productivity improvements will ordinarily have a stronger factual story than a business that simply asserts that its prompt is valuable. Courts applying Michigan law examine factors such as the value of information to the owner and competitors, the resources expended in developing the information, the extent to which the information is known inside and outside the business, the measures used to protect it, and the difficulty with which others could properly acquire or duplicate it.⁵ Those concepts translate naturally to modern prompt development.
The weakness of an individual-prompt claim is that some prompts may be readily reproducible. If a competent user could independently write substantially the same prompt in a few minutes using ordinary knowledge, the prompt may lack the required secrecy-based economic value. Trade-secret law does not prevent competitors from independently developing their own solutions. Nor does it give one business ownership over general prompting principles that are already widely known. The stronger case therefore usually involves evidence that the protected information is not simply an obvious instruction, but a refined and nonpublic solution that confers a real competitive advantage.
Prompt libraries may be more naturally suited to trade-secret protection than isolated prompts because Michigan’s statutory definition expressly includes a “compilation.”¹ A business might maintain hundreds of prompts organized by business function, customer type, product, jurisdiction, transaction, technical issue, or operational scenario. The individual prompts may vary in sophistication, but the library as a whole can reflect substantial investment in identifying useful use cases, refining instructions, documenting successful approaches, selecting models, incorporating examples, and establishing quality-control procedures.
The value of such a library may lie precisely in the compilation. A competitor could theoretically create its own prompts one at a time, but obtaining an established library could eliminate months or years of experimentation. It could reveal not simply how the company uses AI, but which problems the company has learned to solve with AI, which approaches failed, which prompts work reliably, which variations are appropriate in particular circumstances, and which internal knowledge is important to producing acceptable results.
The treatment of combined information in Mike’s Train House is useful by analogy. The court rejected the proposition that information necessarily loses trade-secret protection merely because some of its components are known.⁴ Similarly, InteliClear, LLC v. ETC Global Holdings, Inc., although arising outside Michigan and therefore persuasive rather than binding authority, illustrates how proprietary value can exist in the structure and interrelationship of software-related information. The Ninth Circuit concluded that uniquely designed database structures, processes, and combinations could support trade-secret protection when described with adequate specificity.⁸ A carefully maintained prompt library can raise a comparable argument: the protectable information may consist of the organized system and relationships among components rather than every individual word viewed separately.
This distinction may become especially important where a former employee exports an entire prompt repository shortly before leaving for a competitor. The owner does not necessarily need to contend that every line in every prompt is unknown to the world. The more persuasive theory may be that the employee acquired a curated, tested, internally organized body of information that took substantial resources to develop and that would provide a competitor with an immediate shortcut.
The term “prompt” can also understate what businesses are actually developing. Sophisticated AI deployments increasingly involve workflows rather than one-time questions to a chatbot. A workflow may instruct one model to classify incoming information, another process to retrieve selected internal data, a second prompt to produce an initial analysis, an automated rule to identify certain risk conditions, and a human reviewer to validate the result before a final output is generated. The workflow may also include proprietary templates, internal reference materials, model-selection decisions, routing logic, quality thresholds, escalation rules, and validation procedures.
Michigan’s statute expressly extends beyond compilations to methods, techniques, processes, and programs. ¹ That language makes trade-secret law a particularly plausible form of protection for confidential AI workflows. The valuable secret might not be the text of Prompt A or Prompt B. It may instead be the knowledge that Prompt A should be used first, that its output should be transformed in a particular way, that a specific internal dataset should then be consulted, that particular exceptions must be tested, and that a second model performs best for a later stage of the process.
A well-designed workflow can effectively capture organizational know-how. A business may have learned through years of experience which information matters, which warning signs deserve scrutiny, which questions should be asked, which mistakes frequently occur, and which decisions require human approval. Translating that institutional knowledge into an AI workflow does not necessarily make the knowledge public. To the contrary, the resulting system can make it easier to identify and document exactly what the organization regards as proprietary.
This is also consistent with older Michigan trade-secret principles. Long before generative AI, the Michigan Supreme Court in Hayes-Albion Corp. v. Kuberski considered confidential manufacturing methods and processes and emphasized the distinction between an employee’s general skill and knowledge and an employer’s protected trade secrets.⁷ The technology has changed dramatically, but the underlying legal problem has not. Businesses remain entitled to protect genuinely secret processes while employees remain entitled to use the general experience and skills they acquire in their profession.
The most carefully engineered prompt in the world will not remain a trade secret if the owner does not treat it as one. Under MUTSA, information must be “the subject of efforts that are reasonable under the circumstances to maintain its secrecy. ¹ This requirement turns trade-secret protection into something businesses must actively maintain rather than a status automatically created by inventing something valuable.
For AI assets, the practical consequences are significant. A company that stores proprietary prompts in a publicly accessible shared drive, allows unrestricted copying, provides access to employees who have no business reason to use them, publishes examples containing the essential instructions, or permits workers to transfer the prompts into personal accounts without restriction may have difficulty proving that it genuinely attempted to preserve secrecy. Conversely, password protection, role-based access, confidentiality agreements, access logging, restrictions on exporting repositories, employee training, vendor controls, and documented policies regarding confidential AI inputs can help demonstrate reasonable measures.
The Michigan Court of Appeals’ unpublished decision in Endoscopy Corporation of America v. Kanaan provides a useful modern illustration of the role confidentiality measures can play. The court concluded at the pleading stage that allegations concerning a confidential process, combined with employment obligations requiring confidentiality and the return or destruction of confidential information, were sufficient to support a trade-secret theory. ¹⁰ Although an unpublished decision has limited precedential value, the reasoning reinforces a basic lesson that applies directly to AI systems: contractual restrictions and operational controls can provide evidence that information was actually treated as confidential.
Companies should therefore think of prompt repositories much as they would source-code repositories, pricing databases, engineering specifications, or proprietary operating manuals. The question is not whether every employee must be excluded. Reasonable secrecy frequently requires access by people who need the information to do their jobs. The important point is whether access is controlled in a manner consistent with the information’s asserted value and confidentiality.
The use of third-party generative AI services complicates the analysis because confidential information may leave the company’s direct technical environment. Employees sometimes paste prompts, source material, customer information, internal strategies, code, or other proprietary data into external systems without first considering the contractual and technical consequences. From a trade-secret perspective, the critical questions include what the service provider may do with submitted data, whether it retains prompts and outputs, who may access them, how long they are retained, whether the data may be used for model improvement, and what enterprise controls are available.
The relevant lesson is not that using a third-party AI service automatically destroys trade-secret protection. Reasonableness depends on circumstances. Businesses routinely entrust confidential information to outside lawyers, accountants, cloud providers, consultants, and vendors under appropriate safeguards. An AI provider can be analyzed in the same way. The legal problem arises when a company claims extremely valuable information is secret while allowing employees to disclose it to third parties under terms or practices inconsistent with confidentiality.
This makes AI governance part of trade-secret governance. A business that wishes to claim prompt libraries and AI workflows as proprietary should decide which AI platforms may receive confidential information, what types of information may be entered, which account configurations are required, and how prompt histories and outputs are retained. Those decisions should not remain an informal understanding among technically sophisticated employees. Documented practices make the later argument for “reasonable efforts” substantially more credible.
Trade-secret protection cannot simply become a mechanism for preventing employees from using the general skills they acquired while working for a company. Michigan authority reflects a strong distinction between protected confidential information and an individual’s general knowledge and experience. Hayes-Albion recognized both the legitimate protection of trade secrets and the public interest in allowing employees to pursue their occupations.⁷ The Sixth Circuit similarly explained in Degussa Admixtures, Inc. v. Burnett that an employee’s general knowledge is not itself a trade secret.⁶
That distinction is particularly important in the field of AI because an employee who develops sophisticated prompts will inevitably become better at prompting. The employer may have a legitimate claim to a confidential prompt repository, a particular workflow, proprietary testing results, or nonpublic instructions developed on company time. It ordinarily should not be able to convert general expertise such as “knowing how to prompt a large language model effectively” into a trade secret.
Michigan also does not simply presume misappropriation because an employee with access to confidential information goes to work for a competitor. In CMI International, Inc. v. Intermet International Corp., the Michigan Court of Appeals rejected an expansive inevitable-disclosure theory and explained that generalized allegations about trade secrets and a competitor’s employment of a knowledgeable former employee are not enough. ³ This principle has practical consequences for AI disputes. A company should be prepared to identify the particular prompt library, workflow, model configuration, evaluation dataset, or other information allegedly taken and to establish evidence of acquisition, disclosure, use, or threatened misuse rather than relying solely on the employee’s new position.
For employers, this makes onboarding and offboarding especially important. When employees create AI resources as part of their work, ownership and confidentiality expectations should be clear. When employees depart, the company should know where prompt repositories are stored, whether copies have been downloaded, whether personal AI accounts were used, and whether access credentials have been revoked. A well-documented offboarding process can both prevent actual loss and create evidence if a dispute later occurs.
A company cannot ordinarily prevail by alleging that its “AI know-how” was stolen. Courts need to understand what the asserted trade secret actually is. Dura Global Technologies, Inc. v. Magna Donnelly Corp. is frequently cited for the requirement that a claimant particularize and identify the material claimed as a trade secret with specificity.⁵ This requirement protects defendants from having to guess what information is supposedly prohibited and allows courts to determine whether the information actually satisfies the statutory definition.
AI disputes can make this identification problem difficult because systems may contain overlapping layers of proprietary and nonproprietary information. The trade secret might be a specific master prompt, a version-controlled prompt library, a chain of prompts connected to particular internal data, an evaluation rubric, a retrieval configuration, an automated decision tree, or the overall combination of those components. The claimant should be able to explain the boundaries of the asserted secret without reducing the allegation to an amorphous category such as “our AI system.”
At the same time, specificity does not necessarily require disclosing every secret publicly in a complaint. Courts have procedures for handling confidential material, and Michigan’s statute expressly allows courts to take reasonable means to preserve the secrecy of alleged trade secrets during litigation. ¹ The practical objective is to define the protected information precisely enough for the court and opposing party to evaluate the claim while avoiding unnecessary public disclosure.
The reasoning in InteliClear is again instructive by analogy. There, the plaintiff survived a challenge because it identified particular aspects of its database’s logic, tables, structures, processes, and interrelationships rather than merely pointing to its entire software product and asking the court to determine which parts were secret.⁸ Businesses developing AI systems can learn from that approach. Maintaining documentation of prompt versions, workflow diagrams, testing protocols, performance results, access rights, and development history may later make it much easier to describe what the company actually claims as proprietary.
Not every useful AI prompt has independent economic value from being secret. That distinction should receive careful attention. A prompt may be extremely useful yet widely available. A company may use a clever workflow that has already been described in vendor documentation, public articles, online repositories, conference presentations, or widely circulated training materials. The fact that the business benefits from the workflow does not establish that the benefit results from secrecy.
A stronger trade-secret case exists where possession of the information provides a meaningful advantage that another business could not readily obtain through proper means. A proprietary prompt library could reduce development time. An internal AI workflow could encode years of lessons about customer behavior or manufacturing failures. A confidential evaluation framework could reveal what the company considers a successful output. A collection of model-specific prompting techniques could substantially improve reliability in an application where errors are costly. In each case, the owner should be able to connect confidentiality to competitive value.
Evidence can include development costs, testing history, efficiency improvements, reduced error rates, improved conversion rates, faster turnaround, avoided labor expense, increased capacity, or the difficulty a competitor would experience recreating the system independently. Trade-secret protection becomes more persuasive when the company can explain not simply that “we worked hard on this,” but why obtaining the information without undertaking that work would provide another party with an unfair commercial shortcut.
Even when a prompt or workflow qualifies as a trade secret, trade-secret status alone does not establish liability. MUTSA separately defines misappropriation to include acquisition by improper means and certain unauthorized disclosures or uses by persons who know or have reason to know that the information was obtained through improper means or under circumstances imposing a duty of secrecy. ¹ Theft, misrepresentation, breach of a duty to maintain secrecy, and electronic espionage are among the types of conduct contemplated by the statute.
An employee who downloads an internal prompt repository before joining a competitor presents a different case from a competitor that independently develops similar prompts. A consultant who was given access to a confidential workflow under a nondisclosure agreement and later provides it to another customer presents a different case from a customer who observes an externally visible product feature and develops its own method. Trade-secret law protects against misappropriation, not legitimate independent development.
This distinction will become increasingly important because different organizations using the same underlying AI models may independently arrive at similar prompting techniques. Similarity alone does not necessarily establish copying. Businesses asserting trade-secret claims should preserve evidence concerning repository access, downloads, file transfers, version history, cloud activity, emails, messages, system logs, and subsequent use. Strong digital evidence can transform a speculative suspicion into a factually grounded claim.
If a protected AI asset is misappropriated, Michigan law provides meaningful remedies. MUTSA permits injunctions against actual or threatened misappropriation and authorizes courts in appropriate circumstances to compel affirmative acts designed to protect the secret. ¹ The statute also permits recovery for actual loss and unjust enrichment that is not otherwise included in the loss calculation, and in some situations, damages may be measured by a reasonable royalty. ¹ Attorney fees may be available where willful and malicious misappropriation exists, while a claimant that brings a misappropriation claim in bad faith can itself face a potential fee award. ¹
The federal Defend Trade Secrets Act may provide an additional cause of action where its interstate-commerce requirements are satisfied. ² Michigan businesses therefore may have both state and federal avenues available depending on the circumstances. The basic substantive questions remain closely related: what information was secret, why did secrecy create economic value, what reasonable measures protected the information, and how was it improperly acquired, disclosed, or used?
The possibility of substantial remedies makes careful claim selection important. A business should resist the temptation to label every AI-related asset a trade secret merely because an employee departed or a competitor introduced a similar system. Degussa illustrates the risks associated with unsupported trade-secret litigation and reinforces that general employee knowledge and legitimate competition remain protected interests.⁶ The strongest claims are built around defined confidential assets, documented secrecy practices, evidence of commercial value, and concrete proof of wrongful conduct.
The most important work usually occurs before a dispute. Trade-secret law rewards businesses that can demonstrate that they recognized valuable confidential information and treated it accordingly. For AI systems, that begins with identifying which assets genuinely warrant protection. A routine prompt available to everyone in the organization may not justify extensive controls. A master prompt that drives an important revenue-producing process, a curated prompt library developed through years of experimentation, or a workflow incorporating confidential customer and operational knowledge may justify significantly stronger safeguards.
The business should then treat those AI assets consistently with their asserted importance. Confidentiality provisions should be broad enough to encompass prompts, workflows, automation logic, model configurations, evaluation methods, and other AI-related information where appropriate. Internal repositories should reflect meaningful access controls. Employees should understand whether they may use personal AI accounts for company work and whether confidential material may be entered into external systems. Vendor agreements should be reviewed with prompt and data confidentiality in mind. Development records should identify who created important assets and how they evolved. Departure procedures should address prompt libraries and AI workspaces just as they address laptops, source code, customer databases, and other proprietary material.
None of these measures alone guarantees trade-secret status. The statutory standard is reasonableness under the circumstances, not perfection. ¹ What matters is whether the overall conduct of the organization is consistent with its later assertion that the information was valuable because others did not know it.
Businesses evaluating their AI intellectual property should avoid focusing too narrowly on whether a single prompt can be “owned.” That framing can miss the most valuable information. In many commercial applications, competitive advantage will come from a system consisting of prompts, data selection, sequencing, model choices, examples, retrieval methods, evaluation standards, automation logic, and human-review procedures. Michigan trade-secret law is sufficiently broad to analyze that system as a compilation, method, technique, process, or program rather than forcing the owner to isolate one supposedly secret sentence. ¹
The combination principle recognized in Mike’s Train House strengthens that conclusion.⁴ Publicly known components do not automatically prevent protection when their confidential combination produces a competitive advantage. AI systems are particularly likely to fit this pattern because most businesses build with technologies that others can also obtain. The underlying model may be commercially available. The vocabulary used in a prompt may be ordinary. The conceptual techniques may be discussed publicly. What competitors may not possess is the particular configuration developed through experimentation, internal expertise, proprietary data, and repeated refinement.
That does not mean every prompt library or workflow is a trade secret. The owner must still establish secrecy, economic value tied to secrecy, reasonable protective measures, and, in litigation, misappropriation of sufficiently identified information. Businesses that publish their prompts, fail to control access, rely entirely on commonplace instructions, or cannot explain why their system would give a competitor an advantage may have difficulty meeting those requirements.
For organizations that make the investment, however, trade-secret law may become one of the most important legal mechanisms for protecting the operational layer of artificial intelligence. Copyright law may present difficult questions concerning short prompts or functional instructions. Patent protection may be unavailable, impractical, expensive, or inconsistent with a company’s desire to avoid public disclosure. Trade-secret protection approaches the issue differently: commercially valuable information can remain protected for as long as it remains secret and continues to satisfy the statutory requirements.
AI prompts, prompt libraries, and AI workflows are not automatically trade secrets under Michigan law, but neither are they categorically excluded from protection. MUTSA’s language is technologically flexible. It protects qualifying compilations, programs, methods, techniques, and processes, and those categories can readily describe many of the proprietary systems businesses are now building around generative AI. ¹
An individual prompt will generally present the strongest case when it reflects substantial nonpublic refinement, produces measurable commercial advantages, is difficult to recreate through proper means, and is actually kept confidential. A prompt library may present an even stronger argument because the collection, organization, testing, and interrelationship of prompts can constitute a valuable proprietary compilation. AI workflows can be stronger still when they encode confidential organizational knowledge through a coordinated process involving prompts, data, routing logic, validation rules, software integrations, and human judgment.
The central lesson is that businesses should not wait until an employee leaves or a competitor launches a similar system to decide whether their AI assets were confidential. Trade-secret protection is built through conduct. Companies that identify valuable AI assets, control access, document development, establish appropriate contractual obligations, govern third-party AI use, and preserve evidence of the assets’ commercial value will be better positioned to invoke Michigan trade-secret law if those assets are later taken or misused.
As AI becomes less of an experimental tool and more of an operating infrastructure, some of the most important intellectual property inside a company may no longer be found solely in its patents, software source code, customer databases, or engineering drawings. It may also be found in the instructions, compilations, workflows, and internal systems that teach commercially available AI technology how to perform the company’s work. Michigan’s existing trade-secret framework provides a credible legal basis for protecting that information when the business can demonstrate that it is genuinely valuable, genuinely secret, and genuinely treated that way.
Contact Tishkoff
Tishkoff PLC specializes in business law and litigation. For inquiries, contact us at www.tish.law/contact/. & check out Tishkoff PLC’s Website (www.Tish.Law/), eBooks (www.Tish.Law/e-books), Blogs (www.Tish.Law/blog) and References (www.Tish.Law/resources).
Footnoted Sources:
1- Michigan Uniform Trade Secrets Act, MCL 445.1901–445.1910, particularly MCL 445.1902(d), MCL 445.1903, MCL 445.1904, and MCL 445.1905. https://www.legislature.mi.gov/Laws/MCL?objectName=mcl-445-1901
2- Defend Trade Secrets Act of 2016, 18 U.S.C. §§ 1836, 1839, particularly 18 U.S.C. § 1839(3). https://content.next.westlaw.com/Glossary/PracticalLaw/Ib8d9eb890fbe11e698dc8b09b4f043e0?transitionType=Default&contextData=%28sc.Default%29
3- CMI International, Inc. v. Intermet International Corp., 251 Mich. App. 125, 132–134; 649 N.W.2d 808 (2002). https://www.casemine.com/judgement/us/59147b5cadd7b0493441b904
4- Mike’s Train House, Inc. v. Lionel, LLC, 472 F.3d 398, 410–411 (6th Cir. 2006). https://law.justia.com/cases/federal/appellate-courts/F3/472/398/473389/
5- Dura Global Technologies, Inc. v. Magna Donnelly Corp., 662 F. Supp. 2d 855, 859 (E.D. Mich. 2009). https://www.casemine.com/judgement/us/59146617add7b04934298ba9
6- Degussa Admixtures, Inc. v. Burnett, 277 F. App’x 530, 534–536 (6th Cir. 2008). https://law.justia.com/cases/federal/appellate-courts/ca6/07-1302/08a0232n-06-2011-02-25.html
7- Hayes-Albion Corp. v. Kuberski, 421 Mich. 170, 181–188; 364 N.W.2d 609 (1984). https://www.govinfo.gov/content/pkg/USCOURTS-mied-4_14-cv-12562/pdf/USCOURTS-mied-4_14-cv-12562-1.pdf
8- InteliClear, LLC v. ETC Global Holdings, Inc., 978 F.3d 653, 657–662 (9th Cir. 2020). https://law.justia.com/cases/federal/appellate-courts/ca9/19-55862/19-55862-2020-10-15.html
9- 3M v. Pribyl, 259 F.3d 587, 595–596 (7th Cir. 2001). https://www.ediscoverylaw.com/2004/12/01/3m-v-pribyl-259-f-3d-587-606-n-5-7th-cir-2001/
10- Endoscopy Corporation of America v. Kanaan, Michigan Court of Appeals No. 359398, slip opinion (March 9, 2023) (unpublished). https://www.courts.michigan.gov/case-search/?page=1&resultType=cases&pageSize=10&aBarNumber=75765