A nine-figure acquisition died on a Tuesday over a verb tense. Not in those words, of course—the term sheet said something more genteel about "unresolved chain-of-title matters relating to founding-era contributors"—but strip away the euphemism and what killed the deal was the difference between agrees to assign and hereby assigns. A single early engineer had signed the weaker phrase, left for a competitor, and stopped returning calls. The buyer's patent counsel found the gap in week three of diligence. The founders found out that the four most valuable words in a software company are not "we built great technology" but "we own it free and clear," and that they had not, in fact, owned it free and clear since year one.

This is the most expensive sentence problem in intellectual property law, and it is entirely preventable. The words that secure clean title to everything a company builds cost nothing to get right at hiring. They cost a fortune to fix later—if they can be fixed at all. This article is about getting them right.

The Defect That Surfaces at the Worst Possible Moment

Meet Lumeneer, a fictional but painfully representative software startup. Four years in, it has a product that works, paying customers, and an acquisition offer worth more than the founders ever let themselves imagine. Then the acquirer's lawyers begin intellectual-property diligence, and they find the problem above: three of the engineers who wrote Lumeneer's core technology in its first year signed invention assignment agreements that used the wrong words—or signed no agreement at all—and the chain of title to the company's most important code is broken. One of those engineers has since decamped to a rival and is not returning calls. The deal does not die outright, but it closes at a haircut, with the founders personally indemnifying the buyer against the title risk they did not know they had created.

We will follow Lumeneer through this article, because its story—with endless variations—repeats across the technology and life-sciences industries with troubling regularity. Companies assume they own the intellectual property their employees create, only to discover, at the precise moments when ownership matters most (a financing round, an acquisition, an IPO, an infringement suit), that defective assignment provisions have left title uncertain. The consequences run from the merely irritating (tracking down former employees for corrective paperwork) to the catastrophic (losing rights to the technology the entire company is built on).

The causes are depressingly predictable. Founders, eager to save on legal fees, download a template from the internet without understanding its limits. Employment lawyers who never took a patent class draft provisions that fail to transfer patent rights. Companies in one state use agreements written for another state's law and trip over statutes they never knew existed. And almost everyone assumes that because the employee drew a salary, the company automatically owns whatever the employee invented—an assumption that is simply, legally, wrong. This article dismantles every one of those traps. For the closely related discipline of protecting what those inventions become, our work on building a trade secret protection program from scratch is the natural companion, and the broader strategy sits within our guide to the legal protection of software. The invention assignment is the foundation under both: without clean title, a trade secret program protects assets the company may not own, and a software IP strategy builds on sand.

The Default Rule That Surprises Everyone: Inventors Own Inventions

Start with the rule that catches founders, executives, and even some lawyers flat-footed: under U.S. law, the default is that the employee owns the inventions they create—even when made on company time, with company equipment, in furtherance of the company's business. This rule, rooted in case law that predates the modern technology economy by a century, reflects a stubborn principle: invention is personal. The inventor, as a constitutional and statutory matter, holds initial title. As the Supreme Court put it in Board of Trustees of the Leland Stanford Junior University v. Roche Molecular Systems, Inc., 563 U.S. 776 (2011), "Since 1790, the patent law has operated on the premise that rights in an invention belong to the inventor." The Patent Act vests the right to obtain a patent in "the inventor," and "unless there is an agreement to the contrary, an employer does not have rights in an invention which is the original conception of the employee alone." United States v. Dubilier Condenser Corp., 289 U.S. 178, 189 (1933).

The practical meaning is stark. Without an effective assignment, the employer holds no ownership of the inventions its employees create, regardless of whether the invention relates to the employer's business, was made during work hours, or used company resources. The salary buys labor; it does not buy the invention. That counterintuitive default is the engine that drives this entire field of law: because the law gives the employer nothing by default, every scrap of ownership the employer enjoys must be affirmatively captured by contract, and every defect in that contract is a hole through which ownership leaks.

There are three principal ways the law softens this default, and it is worth understanding each precisely—including its limits—because two of them are far weaker than employers hope.

The shop right. When an employee uses an employer's time, materials, or facilities to create an invention, the employer may acquire a "shop right"—a non-exclusive, non-transferable, royalty-free license to use the invention. The doctrine is equitable, grounded in fairness and estoppel: having let the employer contribute resources and watched it rely on the invention, the employee cannot then sue the employer for infringement. But notice what a shop right is not. It is not ownership. The employer cannot stop anyone else from practicing the invention, cannot sell or exclusively license it, cannot sue infringers, and cannot transfer the right in a sale of the business. A shop right keeps the employer out of court as a defendant; it does nothing to make the employer the master of the asset. For a company whose value is its exclusive control over its technology, a shop right is a consolation prize.

The hired-to-invent doctrine. Where an employee is hired specifically to invent—to solve a particular problem, or to apply their inventive faculties to a defined assignment—courts will imply an obligation to assign the resulting invention and patent rights to the employer. The leading articulation again comes from Dubilier: "one employed to make an invention, who succeeds, during his term of service, in accomplishing that task, is bound to assign to his employer the patent obtained." 289 U.S. at 187. This is genuine ownership, not a mere license. But the doctrine is narrow and treacherous. It applies only when the employee was hired to invent the specific thing (or to solve the specific problem) at issue—not merely to do technical or research work in general. The factual inquiry is unpredictable, fact-bound, and litigated case by case, which is precisely why no competent counsel relies on it. It is a doctrine for the company that forgot to get an assignment, not a substitute for getting one.

Express written assignment. This is the reliable path, and the only one a prudent company depends on. Employees can—and at any well-run company, do—assign their invention rights through a valid written agreement. The Patent Act requires patent assignments to be in writing: "patents, or any interest therein, shall be assignable in law by an instrument in writing." 35 U.S.C. § 261. The remaining 90 percent of this article is about how to make that written assignment actually work, because the gap between a written assignment that works and one that does not is where companies lose their crown jewels.

For Lumeneer, the lesson is bracing: salary alone bought it nothing. Its shop right (if any) let it keep using its own code without being sued by its own former engineer—cold comfort to an acquirer that wanted to own the code. Only the right words, on the right paper, signed at the right time, could have transferred title. Three of Lumeneer's first engineers never signed those words.

The Distinction That Decides Cases: "Hereby Assigns" Versus "Agrees to Assign"

If this article teaches one thing, let it be this. There is a decisive, outcome-determinative difference between an assignment of rights and a mere agreement to assign them, and getting it wrong can cost a company its title to everything its early employees built.

A present assignment—"Employee hereby assigns to Company all right, title, and interest in and to any Inventions"—operates as an immediate transfer. For inventions that already exist at signing, it conveys them then and there. For future inventions, the Federal Circuit treats this language as automatically transferring rights the moment the invention comes into being, with no further paperwork required. An agreement to assign—"Employee agrees to assign"—creates only a contractual promise to make an assignment later. Title does not move until the employee signs a separate assignment document. If the employee never signs it, signs an inconsistent assignment to someone else first, or simply refuses, the employer is left holding a breach-of-contract claim against an individual, not the invention itself.

The Federal Circuit drew this line in stark terms in FilmTec Corp. v. Allied-Signal Inc., 939 F.2d 1568 (Fed. Cir. 1991), the case that made "hereby assign" magic words. There, an inventor had agreed to "grant and assign" future inventions to his employer (a government contractor) and later assigned the same invention to FilmTec. The court held that the operative phrasing mattered enormously: if an agreement uses present-tense language of conveyance—"does hereby grant and assign"—it effects a present transfer of the inventor's expectant interest in a future invention, so that "no further act would be required once an invention came into being; the transfer of title would occur by operation of law." FilmTec, 939 F.2d at 1573. By contrast, an agreement that the inventor "agrees to assign" or "will assign" conveys nothing until a later instrument is executed. The first assignee to obtain a true present assignment wins, and a later present assignment from someone who has already given away their expectant interest captures nothing—because by then there is nothing left to give.

This is not a hypothetical hair-split. It decided Stanford v. Roche. A Stanford researcher signed an agreement stating he "agree[d] to assign" his inventions to the university. Later, visiting a company called Cetus (whose technology Roche acquired), he signed a visitor's confidentiality agreement stating that he "will assign and do[es] hereby assign" his rights to Cetus. When the dispute over a PCR-based HIV-detection patent reached the Supreme Court, the Federal Circuit's application of FilmTec controlled: the researcher's later "do hereby assign" to Cetus was a present assignment that immediately captured his expectant interest, while Stanford's earlier "agree to assign" was a mere promise that had never been executed. Stanford, which believed it owned the invention, owned nothing assignable; it had only a contract claim, and it lost on standing. The Supreme Court declined to disturb the FilmTec rule and affirmed. 563 U.S. at 786–87.

The lesson is unambiguous and free to implement: always use present-assignment language. "Hereby assigns" transfers rights automatically and beats a later grab; "agrees to assign" leaves the door open to exactly the disaster that befell Stanford—and that Lumeneer's first-year agreements, had they used the weaker phrase, would have invited. Sophisticated patent counsel treat "hereby assigns" as non-negotiable. There is no upside to the weaker phrasing and a catastrophic downside, which is the rarest thing in legal drafting: a free lunch. Take it.

A cautionary footnote from the case law underscores how unforgiving courts are about the exact words. In DDB Technologies, L.L.C. v. MLB Advanced Media, L.P., 517 F.3d 1284 (Fed. Cir. 2008), the Federal Circuit parsed an employment agreement in which the inventor "agrees to and does hereby grant and assign" inventions—and found that this present-assignment language automatically vested title in the employer the instant the inventions were conceived, with the consequence that the inventor never owned the patents he later tried to license to his own company. The court remanded to determine whether the specific inventions fell within the agreement's scope, but the doctrinal point was settled: present-tense conveyance language is self-executing, and the inventor's later dealings could not undo a transfer that had already happened by operation of law. Whether "hereby assigns" helps or hurts a given party depends entirely on which side of the assignment you sit; what never changes is that the magic words control.

How the Magic Words Actually Work Over Time

Because the "hereby assigns" rule is so consequential, it is worth seeing exactly how it operates, because the mechanics are genuinely strange. The puzzle: how can an employee assign, today, rights in an invention that does not yet exist and may never exist?

The answer is that a present assignment of future inventions operates as what courts call an automatic, or self-executing, assignment. The operative language—"Employee hereby assigns to Company all right, title, and interest in any Invention conceived during employment"—is treated as a present transfer of the employee's expectant interest. The employee owns an inchoate, future interest in inventions not yet made, and the contract conveys that future interest now. So when the future invention is later conceived, title vests in the company automatically, in the same instant, by operation of law, without anyone signing anything further. FilmTec and DDB Technologies both rest on this conception. "Agrees to assign" promises a future act that must still be performed; "hereby assigns" performs the act in advance, leaving nothing to be done when the invention arrives.

Trace it through Lumeneer. Suppose one of its first engineers, call her Priya, signed a present-assignment agreement on her first day. Eighteen months later, Priya conceives the company's core algorithm. At the moment of conception, title to that invention passes to Lumeneer automatically. Priya never has to be asked to sign anything; even if she refused, the rights would already belong to the company. Now run the counterfactual: Priya signed an "agrees to assign" agreement instead. At conception, she still owns the invention; Lumeneer holds only her promise to assign it later. If Lumeneer remembers to ask, and Priya cooperates, she signs an assignment and title transfers. But if she leaves first, or signs a present assignment to a competitor in the interim, or simply refuses out of spite or leverage, Lumeneer is left suing her for breach while someone else may hold the title.

The two phrasings produce identical results only when everyone cooperates. They diverge precisely in the situations—departures, disputes, competing claims, acquisitions of a defecting employee by a rival—where ownership is contested and therefore matters most. That is the whole point. Contracts are written for the day things go wrong, and on that day, "agrees to assign" is the phrase that fails. This is why a single careless template that says "agrees to assign" can quietly undermine a company's title to everything its early employees built—and why diligence lawyers go word-hunting through founding-era agreements with the grim diligence of a customs inspector.

One important refinement, often misunderstood. Even with present-assignment language that vests title automatically, companies should still obtain a separate, specific written assignment for each patent application—a confirmatory assignment recorded with the USPTO. The two instruments do different jobs and reinforce each other. The present-assignment clause guarantees that the company has title as a matter of law, even if the inventor later becomes unavailable, hostile, or dead. The confirmatory assignment creates the clean documentary record that the Patent Office, future acquirers, and courts want to see in the public chain. Recordation matters in its own right: under 35 U.S.C. § 261, a recorded assignment is valid against a later purchaser if recorded within three months of execution or before the subsequent purchase, so prompt recordation defeats competing transfers and provides constructive notice. Relying on either instrument alone leaves a gap. The automatic assignment without recordation is title nobody can see; recordation without the automatic clause depends on getting a signature you may not be able to get. Pair them.

The Statutes That Tell Employers "No, You May Not Claim That"

An employer cannot simply claim everything an employee ever invents. Several states have enacted statutes protecting employees' rights to inventions that have nothing to do with their jobs—inventions made on the employee's own time, without the employer's resources, and outside the scope of employment. These statutes do something unusual in contract law: they declare certain assignment terms void and unenforceable no matter what the parties signed. Drafting around them is not optional, because consent does not cure them.

The states with such statutes include California, Delaware, Illinois, Kansas, Minnesota, New Jersey, North Carolina, Utah, and Washington. (Nevada is the curious outlier, discussed below, that runs the other way.) While the details vary, the protective statutes share a common architecture: they bar assignment provisions reaching inventions an employee develops entirely on their own time without using the employer's equipment, supplies, facilities, or trade-secret information—except inventions that (a) relate to the employer's business or actual or demonstrably anticipated research or development, or (b) result from the employee's work for the employer. Several also require the employer to give the employee written notice of these protections, and a provision that violates the statute is unenforceable as to the protected inventions.

California is the heavyweight, for the simple reason that so much technology employment happens there. Labor Code § 2870 carves out inventions an employee develops entirely on their own time without company equipment, supplies, facilities, or trade secrets—unless the invention relates, at the time of conception or reduction to practice, to the employer's business or actual or demonstrably anticipated R&D, or results from the employee's work. Section 2871 bars an employer from conditioning employment on a provision purporting to require assignment of what § 2870 protects, but permits the employer to require disclosure of all inventions (held in confidence) so it can determine its rights. Section 2872 requires the employer to give written notice of § 2870's protections in any agreement that requires assignment. The compliance recipe follows directly: a California-facing agreement must exclude § 2870 inventions from its assignment, must include the statutory notice, and must not purport to require assignment of protected inventions. Get it wrong and the offending provision is unenforceable—not voidable, not curable by the employee's signature, but void as written.

Washington (RCW 49.44.140 and 49.44.150) and Minnesota (Minn. Stat. § 181.78) closely parallel California, with two wrinkles worth flagging. Both use the word "directly"—the carve-out applies unless the invention relates directly to the employer's business or anticipated R&D—a narrowing that a careful drafter mirrors in the contractual notice. Both also require the protected invention to be made entirely on the employee's own time without any use of employer equipment, supplies, facilities, or trade secrets, so even minimal use of company resources can knock the invention out of the protected zone and back into the assignable column. Washington additionally requires written notice. The phrase "demonstrably anticipated" carries real weight: to claim an invention as relating to anticipated R&D, the employer must be able to show, with evidence predating the invention, that it actually anticipated work in that area—a hindsight rationalization will not do.

Illinois (765 ILCS 1060/1 to 1060/3) and Kansas (K.S.A. § 44-130) track the same template and likewise require written notice; Kansas adds that the employee must, on request, disclose all inventions being developed, for the purpose of determining rights. Delaware (19 Del. C. § 805), New Jersey (N.J.S.A. § 34:1B-265, the relatively recent "RESTRICT Act" provision), and North Carolina (N.C. Gen. Stat. §§ 66-57.1, 66-57.2) impose the same substantive carve-out but, notably, do not mandate the written notice—though New Jersey and North Carolina expressly permit the employer to require disclosure of all inventions for the purpose of sorting assignable from non-assignable. Utah (Utah Code §§ 34-39-1 to 34-39-3) is structurally distinct and especially demanding: it makes any assignment of a non-"employment invention" created on the employee's own time unenforceable unless (1) employment or continued employment is not conditioned on signing, and (2) the employee receives consideration that is not merely compensation for employment. In other words, in Utah you cannot use the job itself as the price of an over-the-line invention assignment—you must pay separately for it. That is a consideration trap baked into a state statute, and we will return to consideration shortly.

Then there is Nevada (NRS 600.500), the mirror image. Unless the parties agree otherwise in writing, Nevada makes the employer the sole owner of any patentable invention or trade secret an employee develops within the course and scope of employment that relates directly to the employee's work—flipping the national default in the employer's favor. Nevada is a useful reminder that "the default rule" is a generalization with at least one statutory exception, and that a multistate employer cannot assume the inventor-owns-it baseline holds in every jurisdiction.

Most states—including New York, Texas, Massachusetts, Florida, and Georgia—have no invention-assignment statute at all, leaving employers broader latitude. But "no statute" is not "anything goes." Even in those states, an overbroad provision claiming every invention regardless of any connection to employment may be attacked as unconscionable, void against public policy, or (if extracted mid-employment) lacking consideration. The prudent drafter therefore includes reasonable carve-outs even where no statute compels them, both to demonstrate fairness to a skeptical judge and to head off those challenges. For a multistate employer like Lumeneer, the safe course is jurisdiction-appropriate templates rather than a single national form. Lumeneer's downloaded, one-size-fits-nobody agreement is exactly the kind of document that includes a Delaware-flavored carve-out for a California engineer (missing the § 2872 notice, and so vulnerable) or omits the "directly" qualifier its Washington hires were entitled to. The statutory map is not optional reading; it is the difference between an enforceable agreement and a decorative one.

The Consideration Problem Employers Forget

A valid contract requires consideration, and for an invention assignment signed at the start of employment, the job itself supplies it: the employee promises to assign, the employer promises to employ. Clean. The trouble starts when an employer asks existing employees to sign a new or revised agreement. By then the employment relationship already exists, and in several states continued at-will employment, standing alone, is not sufficient consideration for new contractual obligations imposed on a current employee—a principle courts have applied to restrictive covenants in Massachusetts, Pennsylvania, and elsewhere. A related trap is nominal consideration: an agreement reciting "in consideration of $1" without any real exchange may not survive scrutiny, and a recital of consideration that everyone knows was never paid is worse than none.

The fix is straightforward but requires discipline. At hiring, have the employee sign the assignment as a condition of employment, before work begins, so the offer itself is the consideration—the approach the standard Practical Law clauses assume, which is why they recite that no additional consideration is owed. For existing employees, provide genuine independent consideration: a signing bonus, a real raise or promotion, an equity grant, or newly extended access to confidential information or responsibilities—and document the exchange explicitly in the agreement's recitals. The rules vary by state and a multistate employer should plan to the strictest. Texas generally accepts continued at-will employment as consideration if the employee receives confidential information or specialized training in exchange. New York generally accepts continued employment if the employee remains for a substantial period after signing. Massachusetts has historically demanded fresh, independent consideration. And Utah, as we saw, requires by statute that an over-the-line invention assignment be supported by consideration "that is not compensation for employment." A multistate employer like Lumeneer should assume the most demanding standard applies everywhere and provide explicit consideration whenever it asks an existing employee to sign—a small, documented cost that forecloses a later argument that the assignment was never validly made.

The mechanics of a mid-employment rollout deserve real attention, because this is where companies most often manufacture the very defect they are trying to cure. The classic scenario: a startup that used a flawed or missing agreement in year one later hires competent counsel, who prepares a proper agreement and circulates it to the founding team and early employees with a cheerful note asking them to sign "to update our paperwork." If that is all the company does, it may collect a stack of signatures that are unenforceable for lack of consideration—leaving the company no better off than before, and arguably worse, because it now believes a problem is fixed that is not. The diligence lawyers will not be fooled by a 2026-dated signature on a document that gave the signer nothing.

The disciplined approach pairs the new agreement with genuine consideration documented in the agreement itself—a modest signing bonus, an equity grant, a contemporaneous raise or promotion, with a recital that explains the exchange. And where the company is also trying to cure a past defect—an early employee whose original assignment was ineffective or nonexistent—the new agreement should include a present assignment of inventions already made, not just future ones, so that the confirmatory transfer reaches the work already done. "Employee hereby assigns all Inventions, whether conceived before or during employment" is the kind of phrasing that retroactively closes the gap (subject, of course, to the state carve-outs). Handled carelessly, a remediation effort papers over a title gap without closing it. Handled properly—real consideration, present-tense conveyance reaching past and future inventions alike—it converts an uncertain chain of title into a clean one. For a company heading toward a financing or a sale, getting this rollout right is often the single most valuable piece of pre-transaction housekeeping it can do, and the cheapest insurance it will ever buy. Pair this exercise with the disclosure discipline described in how to prepare an invention disclosure for your patent attorney, so that the company actually knows which inventions exist and need capturing.

Defining "Inventions" Broadly Enough—But Not Too Broadly

Two drafting details quietly determine whether an otherwise sound agreement actually delivers ownership, and both are easy to underinvest in. The first is the definition of "Inventions" itself.

The definition must be broad enough to capture everything of value, yet not so sloppy that it collides with the state statutes above or invites a court to find ambiguity and resolve it against the employer. A serviceable definition reaches inventions, discoveries, developments, improvements, designs, works of authorship, know-how, mask works, trade secrets, processes, techniques, methods, ideas, and concepts—whether or not patentable or copyrightable—conceived, developed, or reduced to practice during employment within the defined scope, "regardless of whose equipment or resources were used."

Each omission in that list is a documented way companies have lost rights. An agreement that assigns only "patentable inventions" leaves unpatented but valuable know-how unassigned. One that covers "inventions" but not "works of authorship" may fail to reach software copyrights that do not qualify as works made for hire. One that says "conceived" but not "reduced to practice" may miss inventions developed in stages where the inventive contribution and the working embodiment arrive at different moments. And—the cautionary tale that every IP drafter knows—an agreement that lists "inventions" but neglects the humble word ideas can leave a multimillion-dollar concept unassigned. In Mattel, Inc. v. MGA Entertainment, Inc., 616 F.3d 904, 909–10 (9th Cir. 2010), the Ninth Circuit (Kozinski, C.J.) held that an employee invention agreement's failure to specifically include "ideas" created genuine ambiguity about whether a designer's idea for the Bratz dolls—conceived while he worked for Mattel—was assigned to Mattel at all. A definition that had simply added one word would have spared years of litigation and a reversed quarter-billion-dollar judgment. The lesson is to enumerate generously and assume a hostile reader.

For a software company like Lumeneer, the definition should expressly span source code, object code, algorithms, architectures, data structures, models, and the full range of IP types, so that nothing important falls into a definitional crack between "invention" and "work of authorship." Breadth must then be reconciled with the statutory carve-outs: a California agreement applies its broad definition but expressly excludes anything qualifying under § 2870, so that the definition is comprehensive as to what the company may claim while respecting what the law says it may not. The art is in writing a clause that is maximalist within the legal boundary and silent past it—reaching for everything the law allows and not one inch further, because the inch further is what makes the whole clause vulnerable.

The Cooperation Machinery and the Power of Attorney That Saves the Deal

The second quietly decisive detail is the cooperation-and-power-of-attorney machinery that makes an assignment usable. Owning an invention is not the same as being able to prosecute a patent on it. Patent prosecution requires the inventor's participation—signing the inventor's oath or declaration (which the America Invents Act, at 35 U.S.C. § 115, lets the inventor execute as part of an assignment document), assisting the patent attorney, executing formal papers, and sometimes giving testimony in litigation. An inventor who has left, gone hostile, or simply stopped answering email can gum up the works even when the company indisputably owns the invention.

A well-drafted agreement therefore obligates the employee to cooperate, during and after employment, in securing and perfecting the company's rights: to disclose inventions, to keep adequate records, to sign whatever documents are needed, and to assist in prosecution and enforcement, with the company bearing the cost. (Covering the former employee's expenses is not just fair; it materially improves the quality of cooperation you actually get from someone with no remaining stake.) Because employees nonetheless become unavailable—the recurring villain in chain-of-title problems—the agreement should also include a power of attorney appointing the company as the employee's agent to execute patent and assignment documents if the employee cannot or will not. The well-drafted version makes the power "coupled with an interest" and irrevocable, surviving the employee's later incapacity, so it cannot be revoked precisely when the company needs it.

This single provision is what rescues a company when a departed inventor ignores its calls. Rather than being held hostage—or forced into a specific-performance lawsuit that is slow, expensive, and uncertain—the company executes the necessary paperwork itself, in the inventor's name, and moves on. For Lumeneer, a power of attorney in its early agreements would have meant that the unreachable former engineer's signature was never required in the first place. That one drafting choice would have turned a deal-threatening defect into a non-event. Pair the present assignment, the confirmatory-assignment-and-recordation practice, the cooperation covenant, and the power of attorney, and a company has not just title but the practical ability to exercise it—the difference between owning rights on paper and owning them in fact.

The Copyright Trap: Work Made for Hire Is Not Enough

Here is the gap that catches even sophisticated companies. The work-made-for-hire doctrine and the invention assignment do different jobs, and confusing them blows a hole in the chain of title that no one notices until diligence.

Under 17 U.S.C. § 101, a "work made for hire" is either (1) a work prepared by an employee within the scope of employment, or (2) a work specially commissioned in one of nine enumerated categories where the parties agree in writing that it is a work for hire. For a work made for hire, the employer is the "author" from the moment of creation and owns the copyright (17 U.S.C. § 201(b))—no assignment is necessary or even possible, because the employee never held the rights to assign. The Supreme Court set the test for who counts as an "employee" (versus an independent contractor) in Community for Creative Non-Violence v. Reid, 490 U.S. 730 (1989), importing common-law agency factors: the hiring party's right to control the manner and means of creation, the skill required, the source of tools, the location and duration of the work, the method of payment, the provision of benefits, the tax treatment, and more. The classification is not a label the parties get to choose; it is a multifactor reality test, and a misclassified "contractor" who is really an employee—or vice versa—can upend the analysis.

Now the decisive point: the work-made-for-hire doctrine applies only to copyright. Patent law has no equivalent. There is no doctrine vesting patent ownership in an employer based on the employment relationship alone. An invention made by an employee within the scope of employment, on company time, using company resources, still belongs to the employee-inventor absent an assignment—full stop. This is the trap. A company that owns the copyright in its software as a work made for hire often assumes it therefore owns the patent rights in the inventive functionality that software embodies. It does not. The copyright in the code and the patent on the method the code implements are different assets, governed by different rules, and the work-for-hire doctrine reaches only the former. Without a separate invention assignment, the patent rights remain with the employee.

Because most valuable software contains both copyrightable expression (code, documentation, UI design) and patentable invention (novel functionality, algorithms, methods), a complete agreement needs both a work-made-for-hire provision and a present assignment. This is the "belt and suspenders" structure that good clauses use: a clause stating that copyrightable works within the scope of employment are works made for hire, plus a backup providing that "to the extent any Work Product is not a work made for hire, Employee hereby irrevocably assigns to Employer all right, title, and interest in and to all Work Product and all Intellectual Property Rights therein"—copyright, patent, trade secret, and the rest. The backup assignment catches everything the work-for-hire clause misses: the patent rights it never could reach, plus any copyrights that turn out (because the creator was really a contractor, or the work fell outside the scope of employment) not to qualify as works made for hire after all.

The same belt-and-suspenders logic governs moral rights, the attribution-and-integrity rights that, in the United States, attach to certain works of visual art under the Visual Artists Rights Act and, more expansively, under many foreign laws. Moral rights cannot be assigned, but they can be waived, so a complete agreement includes an express, to-the-extent-permitted-by-law waiver of moral rights in favor of the employer. For a company that may distribute or modify creative work internationally, the waiver is not boilerplate; it is what prevents a former contributor in a moral-rights jurisdiction from objecting to how the company alters or attributes the work.

The contractor scenario is worth dwelling on, because it is where the work-for-hire gap most often produces ugly surprises. A company that engages an outside developer to build a feature, pays the invoice, and assumes it owns the result may own nothing at all. Two doctrines conspire against it. First, a "computer program" is not among the nine categories of specially commissioned works that can be works made for hire, and the contractor is not an employee—so the work-made-for-hire doctrine does not vest the copyright in the company by default; the contractor owns it, absent a written assignment. Second, no work-for-hire concept reaches patents even for employees, let alone contractors. So a company that wants to own what a contractor builds must obtain an express written assignment of all rights—copyright and patent alike—signed by the contractor. Many companies discover this gap only in diligence, when an acquirer asks for the assignment from the freelance developer who wrote a key module three years ago and has since vanished, changed names, or moved to a jurisdiction where serving process is its own ordeal. The fix is trivial at the outset—a signed assignment clause in the contractor agreement, with present-tense conveyance and a power of attorney—and excruciating afterward, which is the recurring moral of this entire subject: the words cost nothing when the work begins and everything when it is too late. The same logic extends to every outside contributor, including open-source maintainers and offshore development shops; for the contract architecture that surrounds these relationships, see our guide to drafting software license agreements.

Holdover Clauses and Prior-Invention Schedules: Two Provisions Pulling in Opposite Directions

Two further provisions round out a well-built agreement, and they cut in opposite directions—one protecting the employer against a departing employee's gamesmanship, the other protecting the employee's pre-existing work from an overreaching employer.

Holdover provisions (also called trailer clauses) extend the assignment obligation for a defined period after employment ends, capturing inventions conceived during employment but not disclosed or reduced to practice until after departure. Their purpose is to close the loophole in which an employee, sensing they are about to leave, sits on an invention conceived on the job and "invents" it formally the week after their last day, claiming it as their own. A reasonable holdover clause says, in effect, "if you actually developed it here, you cannot escape the assignment by waiting to write it down until you have left."

But because holdover clauses restrict a former employee's ability to work and to profit from their own later ingenuity, courts scrutinize them much as they do non-competes, and enforceability turns on reasonableness. The defensible holdover clause shares a family resemblance to a defensible restrictive covenant. It should be as short as the legitimate purpose allows—six months to a year is generally enforceable; beyond two years faces real risk. It should be narrow, limited to inventions actually related to the work the employee performed, not the employer's entire field. It should exclude inventions the employee developed wholly independently of the employer's confidential information. The cleaner versions require the departing employee to disclose qualifying inventions during the holdover period for the employer's evaluation rather than automatically sweeping them in, and some employers provide compensation during the period to bolster enforceability by showing fair exchange.

State law looms large here. Jurisdictions hostile to non-competes scrutinize holdover clauses with the same suspicion, and California—whose Business and Professions Code § 16600 voids most restraints on lawful competition, and whose courts read it expansively—has never definitively blessed holdover clauses, so a clause aimed at California employees must be drafted with particular restraint or avoided. The regulatory backdrop has also been in flux: the Federal Trade Commission's 2024 rule that would have banned most non-competes nationwide did not survive judicial challenge and was ultimately abandoned, returning non-compete enforceability—and, by extension, the climate around holdover clauses—to a state-by-state question. We track that landscape in detail in non-compete agreements under siege: FTC rulemaking and state law developments. The safest holdover clause is one that captures genuinely pre-conceived inventions without functioning as a backdoor non-compete that bars a former employee from working in a field at all.

Prior-invention provisions run the other way, protecting the employee's pre-existing work. They let the employee list, on a schedule attached to the agreement, inventions made before employment that they wish to exclude from the assignment. This humble schedule does several jobs at once: it sets clear expectations about what the employee keeps, it creates contemporaneous evidence of what the employee invented before joining (invaluable if a later ownership dispute erupts), it surfaces potential conflicts with the employer's business before they metastasize, and it implements the carve-outs the state statutes contemplate. It also protects the employer—from an inadvertent claim that it misappropriated the new hire's earlier invention, and from the new hire later improving on a prior invention and claiming the improvement.

The classic failure is the blank schedule. An employee who leaves it empty and later claims a prior invention creates exactly the ambiguity the schedule was meant to eliminate. Best practice forces a choice: the employee must either list prior inventions or affirmatively check "none," with a represent-and-warrant that the list is complete. Employers should watch for overbroad listings (an employee reserving an entire field rather than specific, identified inventions), should acknowledge that the employee need only identify a prior invention—not disclose details that would breach a former employer's confidence—and should require the employee not to incorporate any listed prior invention into company work without written consent. The agreement should also specify what happens if a prior invention is incorporated with consent: typically, a broad, royalty-free, perpetual license to the company to use the prior invention as embedded in the work product, so the company is not later held up by a license the employee can revoke. A representation that the employee is not bound by any conflicting obligation to a former employer rounds out the protection, shielding the new employer from a third party's claim that the hire brought stolen IP through the door.

When Assignments Fail: Standing, Co-Ownership, and the Chain of Title

Defective assignments create gaps in the chain of title—the documented sequence of transfers from inventor to company—and those gaps surface in predictable, high-stakes contexts. Understanding the downstream consequences is what makes the upfront discipline feel worth it.

In M&A diligence, acquirers examine the employment and assignment agreements of every inventor named on every material patent, and verify the chain from inventor to company. Missing or defective assignments delay deals, shave purchase prices, trigger escrows and indemnities, or—when critical ownership cannot be confirmed—kill transactions. This is precisely Lumeneer's predicament. In financing, investors and lenders run IP diligence too; title defects undermine valuation and make IP unusable as collateral, a problem worth understanding alongside the broader financing mechanics covered in navigating the capital raising maze. In patent prosecution, the USPTO requires ownership to be established, and a missing assignment can stall issuance—especially when the needed signature must come from a former employee who is uncooperative or unreachable (the power of attorney, again, is the antidote). And in litigation, standing depends on title.

That last point deserves emphasis, because it is where the doctrine bites hardest. Under 35 U.S.C. § 281, only a "patentee"—the party holding all substantial rights in the patent—may sue for infringement, and a plaintiff without clean title through proper assignments may be dismissed for want of standing no matter how strong its infringement case. Stanford v. Roche is itself, at bottom, a standing case: because Stanford's "agree to assign" language never conveyed title, Stanford lacked the ownership necessary to sue, and a meritorious-looking infringement theory evaporated on a threshold question. A company can be entirely right about infringement and still lose because it cannot prove it owns the patent.

The co-ownership trap is worth making concrete, because its consequences are more severe than most founders expect, and the case law is brutal about it. Suppose Lumeneer's core algorithm was invented jointly by two engineers, and—through an oversight—only one of them signed a valid assignment. The result: Lumeneer and the second engineer are co-owners of the resulting patent. Under default U.S. patent law (35 U.S.C. § 262), each co-owner holds an undivided interest in the whole patent and "may make, use, offer to sell, or sell the patented invention . . . without the consent of and without accounting to the other owners." So the second engineer could license the patent to Lumeneer's fiercest competitor, or practice it himself, without Lumeneer's consent and without owing Lumeneer a penny. Worse still for enforcement: as a rule, all co-owners must join as plaintiffs to sue for infringement, so a co-owner who refuses to participate can prevent Lumeneer from enforcing the patent at all.

The Federal Circuit drove this home in Advanced Video Technologies LLC v. HTC Corp., 879 F.3d 1314 (Fed. Cir. 2018). There, one of three co-inventors had signed an employment agreement saying she "will assign" her rights—future-tense, not present—and never executed a later assignment. Because the agreement was a mere promise to assign rather than a present conveyance, she retained her co-ownership interest. She refused to join the infringement suit, and the court held that the plaintiff, lacking the participation of a co-owner who could not be involuntarily joined, could not maintain the action. The patent was, for litigation purposes, unenforceable—undone by a single inventor and a single verb tense. A missing or defective assignment from one of several joint inventors thus does not merely create a gap; it hands a former employee both a competing license to sell and a veto over enforcement. This is why diligence lawyers treat any inventor without a clean present assignment as a serious problem rather than a paperwork formality, and why the cost of fixing it after the fact—buying out a co-owner who now grasps his leverage—can dwarf the cost of getting the assignment right at hiring.

Other failure modes round out the catalog. Partial assignments (U.S. rights but not foreign, patents but not trade secrets, present inventions but not improvements) fragment ownership and complicate global protection. Competing assignments, as in FilmTec and Stanford, leave the loser holding only a breach-of-contract claim against an individual—often judgment-proof or vanished—rather than the asset. Remediation exists but is uniformly unpleasant: corrective assignments from cooperative former employees (cheap and easy when they cooperate, costly when they do not); cash settlements to obtain reluctant signatures; specific-performance litigation (slow, expensive, uncertain); or, in deals, representations and indemnities that shift the risk to a seller or escrow but do not cure it. The far better course is prevention through process—jurisdiction-appropriate templates reviewed by IP counsel (not employment counsel alone), agreements signed before work begins, prompt USPTO recordation, periodic audits of personnel files, and immediate remediation of assignment gaps when acquiring companies or onboarding teams via acqui-hire.

A Practical Cadence Across the Employment Life Cycle

The disciplines above translate into a rhythm that tracks the employment life cycle, and a company that follows it rarely faces Lumeneer's surprise. The point of a cadence is that ownership stops being a thing you remember to handle and becomes a thing the process handles for you.

Before extending an offer, determine which state's law governs, select the right jurisdiction-specific template, confirm it includes any required statutory notice (California, Illinois, Kansas, Minnesota, and Washington all require it), and review whatever prior inventions and prior-employer restrictive covenants the candidate discloses. In drafting, use present-assignment ("hereby assigns") language; include the state-law notice where required; exclude statutorily protected inventions with the correct qualifiers ("at the time of conception or reduction to practice" for California; "directly" for Minnesota, Kansas, and Washington); define "Inventions" broadly across all IP types, remembering to include "ideas"; pair a work-made-for-hire clause with a backup present assignment; attach a prior-inventions schedule that forces a "list or none" choice; add a reasonable, disclosure-based holdover clause only if genuinely needed and only where state law tolerates it; confirm adequate consideration (especially for existing employees and always in Utah); and include a cooperation covenant, an irrevocable power of attorney, and a moral-rights waiver.

At onboarding, make sure the agreement is signed before the first day of work, retain the prior-inventions disclosure in the personnel file, and document that the employee received the required statutory notices. During employment, maintain a working invention-disclosure process so inventions are actually captured rather than forgotten; obtain a specific confirmatory assignment for each patent application and record it with the USPTO within the three-month window; update agreements when an employee relocates to a different state (a New York hire who moves to California now sits under § 2870); and audit files periodically so gaps are found by you, on a quiet afternoon, rather than by an acquirer's counsel under deal pressure. At separation, confirm all assignments are executed and recorded, conduct an exit interview that asks specifically about undisclosed or in-progress inventions, and document the departure date for any holdover calculation. None of this is expensive. All of it is far cheaper than discovering a title defect when a deal or a lawsuit depends on clear ownership. This onboarding and disclosure discipline dovetails with a broader information-security and confidentiality posture; for the confidentiality half of the equation, see drafting enforceable non-disclosure agreements for technology transactions, and for the program that protects the resulting know-how, trade secrets in the age of remote work and cloud computing.

The AI Wrinkle: Capturing What the Human Contributes

One current development reshapes how a forward-looking company should draft today: the inventorship questions raised by generative AI tools, which are moving from theoretical to operational as engineers increasingly build with AI in the loop. Current U.S. law is settled that a patent inventor must be a natural person—the Federal Circuit held in Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022), that an AI system cannot be named as an inventor under the Patent Act, and the Supreme Court declined review. But that clear rule sits alongside evolving USPTO guidance on inventions merely assisted by AI, where the operative question becomes whether a human made a sufficient inventive contribution to qualify as an inventor at all.

For invention-assignment drafting, the practical implication is concrete. The assignment must capture whatever the human employee contributes, however heavily AI-assisted the work, and the definition of "Inventions" should be broad and tool-agnostic—reaching AI-assisted developments without hinging ownership on the unsettled question of how the law will ultimately treat machine contributions. A company should not assume that an AI-assisted invention falls outside its assignment merely because a model did much of the heavy lifting; whatever the human inventor contributes is theirs to assign, and the agreement should be drafted to scoop it up. Keep the language broad, keep it focused on who the human inventor is rather than which instruments they used, and the assignment remains robust regardless of how the inventorship doctrine settles. We examine the ownership and inventorship side of this question in depth in AI-generated inventions: who owns what the machine creates and, from a comparative angle, in artificial intelligence and inventorship: global perspectives on machine contributions to innovation. For Lumeneer, whose engineers increasingly prototype with AI assistance, the lesson is the same as the lesson of this entire article: draft broadly, draft early, and let ownership turn on the human, not the toolchain.

Frequently Asked Questions

Does my company automatically own what my employees invent because I pay them? No. The U.S. default rule is that the inventor owns the invention, even when it is made on company time with company resources. Board of Trustees of the Leland Stanford Junior Univ. v. Roche Molecular Sys., Inc., 563 U.S. 776 (2011). Salary buys labor, not the invention. The employer may have a non-exclusive "shop right" to use an invention made with its resources, and may, in narrow circumstances, claim ownership under the "hired to invent" doctrine—but neither is a reliable substitute for an express written assignment using present-tense ("hereby assigns") language.

What is the difference between "hereby assigns" and "agrees to assign," and does it really matter? It is the most important phrasing choice in the entire agreement. "Hereby assigns" is a present assignment that transfers the employee's expectant interest immediately, so future inventions vest in the company automatically the instant they are conceived, with no further signature required. FilmTec Corp. v. Allied-Signal Inc., 939 F.2d 1568 (Fed. Cir. 1991). "Agrees to assign" is a mere promise to assign later; if the employee leaves, refuses, or assigns to someone else first, the company holds only a breach-of-contract claim, not the invention. Stanford lost a Supreme Court case over exactly this distinction. Always use "hereby assigns."

We're in California. What do we have to do differently? California Labor Code § 2870 makes unenforceable any provision purporting to assign inventions an employee develops entirely on their own time, without company resources, that neither relate to the employer's business or anticipated R&D nor result from the employee's work. Section 2872 requires you to give the employee written notice of these protections in the agreement. So a compliant California agreement must (1) carve out § 2870 inventions, (2) include the statutory notice, and (3) use the "time of conception or reduction to practice" qualifier. California also voids most non-competes (Bus. & Prof. Code § 16600), so holdover/trailer clauses must be drafted with unusual restraint or avoided. Illinois, Kansas, Minnesota, and Washington have similar carve-out and notice regimes.

If our software is a "work made for hire," don't we own everything in it? Only the copyright. Work made for hire is a copyright doctrine (17 U.S.C. § 101); it has no patent equivalent. The patentable invention embodied in your software still belongs to the employee-inventor unless separately assigned. A complete agreement therefore uses a belt-and-suspenders structure: a work-made-for-hire clause plus a present assignment of all IP rights (copyright, patent, trade secret) as a backup. For independent contractors the gap is worse, because a "computer program" is not among the nine categories of specially commissioned works that can be works made for hire—so without a written assignment, the contractor owns even the copyright.

We used a bad template for our first employees. How do we fix it now? Carefully, and with consideration. Asking existing employees to sign a corrected agreement "to update paperwork" can produce unenforceable signatures for lack of consideration, especially in states like Massachusetts (and always in Utah, by statute, for over-the-line assignments). Pair the new agreement with genuine, documented consideration (a bonus, equity, raise, or promotion), and include a present assignment reaching inventions already made, not just future ones, so the fix actually cures past gaps. Then obtain confirmatory assignments for any issued patents or pending applications and record them with the USPTO. Done right, this is the single most valuable pre-financing housekeeping a startup can do.

What happens if just one of several joint inventors never signs an assignment? Potentially disaster. That inventor becomes a co-owner of the whole patent and—under 35 U.S.C. § 262—can independently license or practice it without your consent and without accounting to you. Worse, as a rule all co-owners must join to sue for infringement, so a non-joining co-owner can block enforcement entirely. Advanced Video Techs. LLC v. HTC Corp., 879 F.3d 1314 (Fed. Cir. 2018), held an infringement suit could not proceed because a single co-inventor whose agreement said "will assign" (not "hereby assigns") had never assigned and refused to join. One inventor and one verb tense sank the case.

Do we need separate assignments for contractors and offshore developers? Yes—always, in writing, with present-tense conveyance and a power of attorney. Contractors are not employees, so neither the work-made-for-hire doctrine (for most software) nor any patent default gives you their work product by operation of law. Without a signed assignment, the contractor owns what they built, and you may own nothing but an invoice. Foreign jurisdictions add their own assignment, recordation, and moral-rights requirements, so coordinate with local counsel for international contributors.

Conclusion: The Words Are Cheap; the Title Is Not

Invention assignment agreements are foundational documents that determine whether a company owns its most valuable assets, and their defects are insidious precisely because they lie dormant—surfacing only when ownership matters most, by which point remediation may be expensive, difficult, or impossible. The investment in proper drafting and process pays across the entire life of the company: clear title smooths financing, enables enforcement, supports acquisitions, and forecloses ownership disputes, while defective assignments create contingent liabilities that sophisticated counterparties will find and exploit—as Lumeneer's acquirer did, and as the deal that died over a verb tense did before it.

For a startup, establishing proper assignment practices on day one is among the most valuable legal steps its founders can take, and one of the cheapest. For an established company, auditing existing agreements and remediating defects—however uncomfortable—is essential housekeeping that is always less costly before a transaction than during one. And for every employer, the throughline is to work with counsel who understand both employment law and patent law, because the failures in this field almost always come from a drafter who knew one half of the problem and not the other—an employment lawyer who never learned that "agrees to assign" loses, or an IP lawyer who never learned that mid-employment signatures need consideration. The two halves have to meet on the page.

The cheapest moment to secure clear title is at hiring. The most expensive is when a deal or a lawsuit depends on it. Between those two moments lies the entire difference between a company that owns what it built and a company that merely hopes it does. The words that close that gap—hereby assigns, signed before the first day of work, with the right carve-outs and a power of attorney—cost nothing. The title they secure is the company itself.

For help drafting or auditing invention assignment agreements across jurisdictions, contact our intellectual property and technology practice or our employment team.

Related Articles

Selected Authorities

35 U.S.C. §§ 115, 261, 262, 281; 17 U.S.C. §§ 101, 201, 204(a). Board of Trustees of the Leland Stanford Junior University v. Roche Molecular Systems, Inc., 563 U.S. 776 (2011); FilmTec Corp. v. Allied-Signal Inc., 939 F.2d 1568 (Fed. Cir. 1991); DDB Technologies, L.L.C. v. MLB Advanced Media, L.P., 517 F.3d 1284 (Fed. Cir. 2008); Advanced Video Technologies LLC v. HTC Corp., 879 F.3d 1314 (Fed. Cir. 2018); Mattel, Inc. v. MGA Entertainment, Inc., 616 F.3d 904 (9th Cir. 2010); Community for Creative Non-Violence v. Reid, 490 U.S. 730 (1989); United States v. Dubilier Condenser Corp., 289 U.S. 178 (1933); Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022). California Labor Code §§ 2870–2872; California Business and Professions Code § 16600; Washington RCW 49.44.140–.150; Minnesota Stat. § 181.78; Illinois 765 ILCS 1060/1 to 1060/3; Kansas K.S.A. § 44-130; Delaware 19 Del. C. § 805; New Jersey N.J.S.A. § 34:1B-265; North Carolina N.C. Gen. Stat. §§ 66-57.1, 66-57.2; Utah Code §§ 34-39-1 to 34-39-3; Nevada NRS 600.500. 37 C.F.R. §§ 3.21, 3.28, 3.31; MPEP § 301.

This article is for general informational purposes only and does not constitute legal advice, nor does it create an attorney-client relationship. Employment and intellectual-property law vary by jurisdiction and continue to evolve; the discussion here may not reflect the most recent developments, and the sample language described is illustrative rather than a substitute for counsel. Consult qualified employment and IP counsel about your specific circumstances.