Embossed notary seal on a legal document

I came within eleven days of losing an acquisition because of a contractor named Dave. Dave had built our first prototype over six weekends in 2014 for $4,000 and a handshake. Four years later, sitting in diligence for the sale of the company, the buyer's counsel asked a simple question: where's Dave's assignment agreement? There wasn't one. Under the default copyright rules, Dave owned the code that the entire product had grown out of, and we ended up paying him $40,000 for a retroactive signature — a number he arrived at after noticing, quite reasonably, that a seven-figure deal depended on his pen. Dave was polite about it. He did not need to be polite about it. That is the whole topic of this page in one anecdote.

IP assignment is the least glamorous paper in the company and the first thing serious diligence examines. It is also almost free to get right at the beginning and brutally expensive to fix after the fact, because the price of a signature scales with how badly you need it. So let's walk the whole stack: why chain of title matters, the invention assignment agreement every employee and contractor signs, the separate founder assignment, the Schedule A trap, the contractor problem, university IP, open source, trademarks, the foreign-founder wrinkles, and the checklist diligence will actually use against you.

A startup is, legally speaking, a pile of contracts wrapped around some intellectual property. Strip away the pitch deck and the office plants and what an acquirer or investor is buying is the code, the brand, the designs, and the patents — plus the paperwork proving the company actually owns them. That proof is called chain of title: an unbroken line of signed assignments running from every human who ever created anything for the company, into the company.

Break the chain anywhere and the value of the whole pile wobbles. Say your co-founder wrote the core algorithm before incorporation and never assigned it. She owns it. Not the company — her, personally, forever, and if she leaves angry she takes a credible claim to the heart of the product with her. Say a freelancer designed your logo on a gig platform with no assignment clause. He owns it. Say a grad student contributed the key model while on a university grant. The university may own it. Each of these is a stranger — or a friend turned stranger — holding a hostage note against your exit.

And investors know exactly how this plays out, because they've watched it play out. IP left sitting with an individual founder is not a yellow flag; it's a deal killer, the kind that ends a process with a polite email about "timing." I've sat on the founder side of that email. The fund didn't think we were frauds. They thought we were sloppy, and sloppy is unfundable at the exact moment you can least afford it. The fix costs nothing when the relationship is good and everything when it isn't. Sign the paper while everyone still likes each other.

The workhorse document goes by a few names — invention assignment agreement, IAA, PIIA (proprietary information and invention assignment agreement) — and every employee and every contractor signs one, on or before their first day of work. Not in week three. Not "when HR gets to it." Day one, before the laptop gets unlocked, because the agreement has to be in force before the invention exists for the assignment to grab it cleanly.

The document does three jobs. First, confidentiality: company information stays in the company. Second, present assignment of future work: the magic words are "I hereby assign," in the present tense — not "I agree to assign," which courts have read as a mere promise that can be broken, and which left one famous litigant arguing ownership of patents years after the inventor had left for a competitor. Present tense, automatic, no further signature needed. Anything the person creates within the scope of their work, during their engagement, belongs to the company the moment it exists. Third, disclosure: the person must tell the company about inventions as they make them, so nobody surfaces at exit claiming the company's product was secretly their side project.

Scope matters, and this is where founders should read instead of skim. A good IAA covers inventions that relate to the company's actual or demonstrably anticipated business, or that result from work performed for the company, or that use company time, equipment, or information. It does not — and in California, under Labor Code Section 2870, cannot — grab everything an employee invents on their own time with their own stuff, unrelated to the business. California, and a handful of states with similar statutes, voids overreaching clauses. So an IAA that claims to own your engineer's unrelated weekend novel isn't stronger paper. It's weaker paper with delusions of grandeur. Keep the scope tight and defensible.

Then there's the exception schedule — usually labeled Schedule A or Exhibit A — where the signer lists their prior inventions. That little blank is important enough to get its own section below. And a practice note: collect these signatures like a paranoid. Every employee, every contractor, every intern, every advisor who touched the product. The set has to be complete, because diligence checks completeness — they take your headcount list and ask for one signed agreement per name. A 96% collection rate is a conversation. Have zero conversations.

Founders need two assignments, and conflating them is the classic self-lawyered mistake. The IAA covers work created during the employment relationship. But at the moment of formation, a founder is not yet an employee of anything — the company is thirty seconds old, no employment agreement exists, and more importantly, the most valuable IP often predates the company entirely: the prototype coded on nights and weekends, the pitch deck, the domain name, the patent application filed from a dorm room, the trained model, the brand.

So founders sign a separate founder intellectual property assignment agreement — often folded into the stock purchase paperwork — that transfers all of that pre-incorporation work to the company as of its creation. This is also, conveniently, part of the consideration for founder stock: you're buying your shares partly with cash at par value and partly with the assignment of everything you built before the company existed. That's why the purchase price can be a penny per hundred shares without the IRS or a future acquirer blinking — the real payment is the IP.

Be specific in this document. Generic language — "all intellectual property related to the business" — invites a fight five years later about what was related. List the assets: the repositories, the domains, the filing numbers of any provisional patents, the names of the apps, the design files. Attach the list as an exhibit. When the acquirer's counsel asks "does the company own the core model," the answer you want is an exhibit with the model's name on it, signed, dated before the first line of post-incorporation code.

One more founder-specific trap: if any founder built any piece of the product while employed somewhere else — and someone always did — read their old employment agreement before assigning anything. Most BigCo IAAs claim after-hours work related to the employer's business. Building your startup in the same industry as your day job, on your own laptop but on their problem space, is how founders end up with a former employer holding a colorable claim to the company. Get a written release or carve-out if there's real overlap. Yes, it's an awkward letter to send. It's less awkward than the one their lawyer sends your acquirer.

The prior invention disclosure schedule — Schedule A — is where each signer lists inventions they made before joining that they want to keep. The logic is brutal and simple: if it's not listed, the company owns it. Everything you invented before this job that you don't write down gets swept into the assignment, or at minimum into a gray zone where proving it was "prior" becomes your expensive hobby.

For founders, this is not boilerplate. It's the document that protects your side projects. Say you built an indie iOS game two years before the startup — a real, shipped thing with real revenue. If you don't list it on Schedule A when you sign your founder paperwork, and the company later gets acquired, nothing in the assignment paper clearly excludes it. Will the acquirer claim your game? Probably not. Do you want "probably" doing that work, in a negotiation where you've already conceded the preference stack? List it. List everything: the game, the old consulting framework, the open source library you maintain, the blog, the abandoned patents. Over-listing costs you an awkward paragraph. Under-listing costs you an asset.

The flip side cuts at your employees, so use it honestly: when a new hire lists "a machine learning recommendation engine" as a prior invention and then the company's product develops a suspiciously similar engine, that Schedule A entry is the paper that decides who owns what. Review every employee's Schedule A at signing — not at the dispute. If an entry describes something uncomfortably close to your roadmap, have the conversation in week one, in writing. And keep the schedules with the signed agreements, because a Schedule A that isn't attached to the agreement it references is a rumor.

Founders coming out of a previous employer in the same space should treat their own Schedule A with the same care — combined with the release discussion above. The list is both shield and map: it shields what's yours, and it maps, for the company's benefit, exactly where the boundary between old work and new work runs. Diligence will read these. Write them like they'll be read, because they will be.

Here's the rule that surprises every first-time founder: employees create IP that generally belongs to the employer under the work-for-hire doctrine. Contractors don't. An independent contractor owns everything they create for you by default, unless two things are true — the work falls into one of the statutory work-for-hire categories and there's a written agreement saying so, or the contractor signs an outright assignment. In practice you want the assignment clause, in writing, signed before work starts, with present-tense "hereby assigns" language, covering everything created in the engagement.

The Daves of the world — my $40,000 lesson — live in this gap. Every startup's early history is a sedimentary layer of contractors: the designer from a gig platform, the overseas dev shop, the friend who "helped with the database." Gig platform terms often include an IP transfer on payment, which is something, but read them — some transfer only on full payment, some carve out the freelancer's pre-existing tools, and dev shops notoriously retain ownership of their internal frameworks and license you the result, which is fine until diligence asks who owns the framework your product can't run without.

And misclassification is the trap with teeth. Call someone a contractor, treat them like an employee — set hours, company laptop, indefinite engagement, managing their daily work — and you've created two problems at once. The tax and labor problem: back payroll taxes, benefits liability, penalties, the works. And the IP problem: the whole engagement sits in a legal fog where their assignment clause, if you even got one, was signed under a mischaracterized relationship, and a departing "contractor" with a grudge and a lawyer can argue about all of it. Classify people honestly. If they work like employees, hire them as employees and get the IAA. If they're genuinely independent, get the contractor agreement with a real assignment clause before the first commit lands in your repo.

The cleanup protocol for legacy contractors: inventory everyone who ever touched the product, match each to a signed agreement, and chase the gaps while the amounts are small. A retroactive assignment signed during a calm month costs a modest thank-you payment. The same signature requested during your Series B diligence costs whatever the signer notices your round is worth. The market rate for a pen, it turns out, is entirely situational.

If any founder or early employee came out of a university — as a student, a grad researcher, a postdoc, a professor — assume the institution has an IP policy that claims inventions made with university resources, under university grants, or within the scope of university employment. That assumption is usually correct. Students often believe "I paid tuition, so I own my work." The IP policy they clicked through at enrollment usually says otherwise, at least for anything touching their research, their lab, or federal grant money.

The Bayh-Dole Act adds federal spice: inventions made with federal funding carry obligations to the government, including a license the government can march in and exercise. None of this is fatal — universities license technology to startups every day, and most tech transfer offices are professional and fast. But the sequencing matters enormously: negotiate the license or the release before you fundraise, not during. A fund asked to close on a company whose core technology is owned by a state university will either walk or price the uncertainty into your round, and the university — which knows exactly when your option term sheet expires — has no incentive to hurry. I've watched a tech transfer negotiation consume five months of a six-month runway. The founders kept the company. Their ownership percentage of it changed meaningfully.

Practical moves: read the actual IP policy of the institution, not the summary on the website. Determine whether the invention disclosure obligation was triggered the moment the thing worked in the lab. If the university disclaims interest, get that in writing — a formal release letter, not a professor's encouraging email. And if you license, fight for an exclusive license in your field of use with the right to sublicense; acquirers discount non-exclusive core technology, correctly, because a competitor can license it too.

Every modern codebase is a layer cake of open source dependencies, and every diligence process now runs a scan — Black Duck or a similar tool — that itemizes every license in your repo. The scan is not the problem. The licenses are.

The permissive licenses are fine. MIT, Apache 2.0, BSD: use it, modify it, sell products built on it, just keep the attribution notices. No diligence team has ever flinched at MIT. The copyleft licenses are where it gets interesting. GPL — especially GPLv2 and GPLv3 — requires that derivative works distributed to others be licensed under the GPL, source code included. Link GPL code into your proprietary product, distribute the product, and a strict reading says your proprietary code becomes GPL too. Your entire moat, open sourced, as a licensing condition. That is a red flag in diligence the way smoke is a red flag in a kitchen. AGPL is stricter still — it triggers on network use, so even a SaaS product that never "distributes" anything gets caught.

The fix is hygiene, not abstinence. Document your dependencies — a simple software bill of materials, updated as you ship. Set a policy: permissive licenses yes, copyleft in the proprietary product no, LGPL and weak copyleft case-by-case with a lawyer's shrug on file. Run the scan yourself once a year, before someone else's counsel runs it for you. And when the scan finds a GPL component — it will, some npm package three levels deep — replace it or isolate it before diligence finds it first. Founders who hand over a clean SBOM get through this section of diligence in an afternoon. Founders who discover their GPL problem mid-process get to re-architect under a closing deadline, which is exactly as fun as it sounds.

The brand follows the same rule as the code: it should belong to the company, not to a founder. If a founder registered the trademark personally — common when the name predates incorporation — assign it to the corporation with a written trademark assignment, and record it. Unrecorded assignments invite chaos later, including the founder's estate arguing about the brand, which is a sentence I wish were hypothetical.

Search the USPTO database before you fall in love with a name, and search it again before you print anything expensive. A federal registration runs a few hundred dollars per class of goods through the USPTO, and the filing itself forces you to think about what you're actually claiming. The diligence angle is simple: an acquirer wants the trademark registered, owned by the company, and free of the threatening letter from the company in another state that's used the name since 2011. That letter surfaces in week two of every deal involving a name nobody searched. Search the name. It's an hour.

Same treatment for domains and social handles, while you're at it — registered to the company, paid from the company account, with the registrar login in the company password vault and not in the personal account of whichever founder set it up at 2 a.m. in 2019. Assets stranded in personal accounts are a chain-of-title problem wearing a trivial disguise.

Cross-border teams hit a wall US founders never see: in several countries, employee-inventor rights are statutory and cannot be signed away, at least not for free. Japan's patent law gives employee inventors a right to "reasonable compensation" for assigned inventions, and courts there have awarded sums that made multinational employers reconsider their bonus structures. Germany's employee invention act is a whole statutory regime — the employer must formally claim the invention, and the inventor is owed statutory remuneration calculated from the invention's value; you don't contract around it, you comply with it. France similarly guarantees employee inventors additional compensation for inventions made under an inventive mission. The pattern: the assignment document can transfer ownership, but the inventor keeps a payment right that no signature kills.

What this means practically. If your co-founder invented the core technology while employed in Munich or Lyon or Tokyo, that prior employer may hold rights your Delaware assignment paper doesn't reach. If you employ engineers in those countries, your US-form IAA is necessary but not sufficient — you need local-law-compliant invention clauses and a compensation mechanism, or you've built a ticking claim into your cap table's foundation. And if a founder brings pre-incorporation IP created abroad, diligence will ask which law governed its creation, because the answer determines whether the assignment actually assigned anything. Local counsel in the inventor's country, for a few hundred dollars, is the cheapest insurance in this entire article.

When a fund or an acquirer diligences your IP, they work from a list that looks remarkably like this one. Build the folder before the first meeting:

Send that folder complete and you change the whole tone of diligence. The other side came looking for leverage — a missing signature here, a stray GPL dependency there, anything to justify "we found some issues, so about that price." Finding nothing is disarming in the best way. And if assembling the list surfaces a gap — and at a real company, it usually surfaces one — fix it now, in the calm, at calm prices. The Dave rate only goes up.

The paper in this guide sits directly on top of the formation paper from our incorporation walkthrough — the founder assignment references the stock purchase, the IAA references the board consent, the whole stack references itself. That's the point. A company is a web of documents that corroborate each other, and diligence is the process of checking whether they do. Make yours boring. Then go read the term sheet guide, because once the IP is clean, the next fight is over what it's all worth.

Venture Legal Hub is not a law firm. This content is for informational purposes only and does not constitute legal advice.