Mission lock for early-stage AI startups: from easiest to hardest
You started your company to make AI go well. You want it to stay that way after the next round, the next board seat and the next acquisition offer. The standard advice is to incorporate a standard Delaware C-corp, sign the standard documents and think about governance later. This post gives you the alternative: eight steps, ordered from cheapest to most expensive, with a clear line between what you should do now and what you should only prepare for.
We wrote it for founders at the stage where most of our portfolio is: pre-seed to Series A, a small team, no CFO, and investors who expect you to move fast. Everything on the “do now” list costs about as much as the standard setup.
This is not legal advice. Every structure below needs a lawyer who has done it before. Use this post to ask that lawyer the right questions.
First: strategy, people, then form
Three things decide whether a mission survives: a good strategy, the right people, and the corporate form. They matter in that order. Beautiful structures with the wrong people fail. Weak structures with great people often do great work. The form is scaffolding around the company. It does not row the boat.
So the ladder starts with no legal work at all.
The ladder at a glance
| Tool | What it protects | Effort | Pre-seed / seed? | |
|---|---|---|---|---|
| 0 | Written values, tied to pay and reporting | Direction of the team | Days | Do now |
| 1 | Public benefit corporation (PBC) | The board’s duty and the right to refuse a bad buyer | Same as a normal corporation | Do now |
| 2 | Founder sentinel share | The PBC itself, against a vote to remove it | A charter clause | Do now |
| 3 | Mission-locked IP license (license in) | Core IP, against use outside the mission | Hours | Do now if you own key IP |
| 4 | Springing public license (license out) | Core IP, if the company drops the mission or fails | Hours to days | Do now, or at the first impact-investor round |
| 5 | Philanthropic money inside the company | Funding for low-margin mission work | A policy and a bank account | Do when a funder asks |
| 6 | Independent public charity next to the company | Mission work that cannot pay for itself | Much of a senior CFO’s time | Prepare only |
| 7 | Perpetual purpose trust | The mission, permanently | Expensive to set up and to run | Prepare only |
0. Write down your values and connect them to something real
Before you ask “PBC or something more exotic?”, write down what you want the company to value when there are other people on the cap table and in the boardroom. Forget the legal shape for now. What should the company choose when the mission and the money disagree?
Values only matter when they are connected to decisions. Connect them to three things:
- Compensation. What behaviour does the company reward?
- Your public benefit (see step 1). The benefit statement should follow from the values.
- Reporting. What do you tell investors and the team each year about how you did?
Without this, every structure below is scaffolding around an empty building.
1. Become a public benefit corporation
Cost: the same as a normal corporation. When: now.
The short version: the difference between a normal corporation and a PBC is the difference between may and shall. A normal company may do many things that do not maximize shareholder value. A PBC’s board must balance three things in its decisions: the stockholders’ financial interests, the interests of the people the company materially affects, and the specific public benefit written in its charter.
Why founders should care
The largest practical value is control in a sale. If a buyer offers twice the price but you believe they will use your technology against your public benefit, the PBC form gives the board a legal basis to say no. Imagine a social network with the public benefit “preserve democratic discourse”. Its board could evaluate a buyer on that basis, not on price alone.
How to draft the public benefit
- Describe the what. Not who you are, why you care, or how you will do it. Those change; the what should not.
- Keep it short and simple. It is the thing the board owes a duty to, in addition to shareholder value.
- Make it measurable. You must report to stockholders on how you pursue it (in Delaware, at least every two years). “Maximize the well-being of all sentient beings” is too broad: you cannot show progress on it each year. The board approves the measurement method.
What enforces it
In Delaware, stockholders who hold at least 2% of the shares can sue to enforce the balancing duty (DGCL §367). The board also must consider stakeholders such as employees and the community, but it does not owe them a fiduciary duty.
The weakness
Since 2020, a Delaware corporation can convert into or out of PBC status with a simple majority stockholder vote (plus any class votes your charter or preferred stock requires). That made it easy for founders to convert. It also makes it easy for a future majority to convert you back. That is why step 2 exists.
Already a C-corp?
You can convert with venture investors on the cap table. The mechanics are a charter amendment approved by a majority of stockholders and any class vote your preferred stock has. The argument to your investors: a PBC protects the company against a hostile or misaligned buyer, and it opens the door to philanthropic capital (step 5) that does not dilute them.
2. Give the mission a sentinel, and make the first sentinel you
Cost: a clause in the charter. When: now.
A PBC needs someone who can block a change to the public benefit or a conversion out. Call this the sentinel. Large organizations spend years designing theirs. For a startup, the sentinel should be the founders first.
The mechanism: create a separate class of stock (a Class B, or “founder preferred”). Write in the charter that the company cannot change its public benefit or convert out of PBC status without the approval of that class. Also allow that stock to be transferred later. When the company is large, the founders can move the sentinel share to a perpetual purpose trust, a public charity or another independent body.
This is the shape Anthropic uses. Its Long-Term Benefit Trust holds a special class of stock with defined powers over the board. You do not need the trust today. You need the share class that a trust can hold later.
3. Hold key IP outside the company and license it in with a mission lock
Cost: hours. When: now, if a founder owns valuable IP.
A license is one of the most effective mission-lock tools available, and one of the least used. A founder who owns valuable IP keeps it outside the corporation. The founder licenses it to the company royalty-free, on the condition that the company uses it for the mission. If the company stops pursuing the mission, the license can end or start to carry a royalty.
Later, the founder usually transfers the IP, or the license agreement, to a charity. From then on it is very hard to take out.
- Valuation: this does not usually hurt valuation. Investors may ask you to move the IP into the company before a sale. That is a negotiation, and you enter it with leverage.
- Worst case: you unwind it.
- The hard part: someone must decide when the company is no longer pursuing the mission. That is the sentinel again (step 2).
- Effort: a model can draft a first version of this license in an hour. Have a lawyer check it before you sign it. The license only protects you if it has real teeth.
4. Add a springing public license on what the company builds
Cost: hours to days. When: now, or when an impact-focused investor joins.
Step 3 protects IP the founder brings in. This step protects IP the company creates. The company grants a license that “springs” into force automatically if the company stops using the technology for the public purpose, or if it goes bankrupt. When it springs, the technology becomes available for public use. The IP cannot be sold off in a fire sale to someone who will use it for the opposite purpose.
This is common in philanthropic investing. The Gates Foundation, for example, attaches global access commitments to its investments, and the same pattern is used in energy and food. For an AI safety company, the natural trigger is: the company stops using the technology for safety.
5. Take philanthropic capital inside the company
Cost: a policy and a separate bank account. When: when a foundation or donor wants to fund you.
More philanthropic money is flowing into AI safety every year. You can accept it without a second legal entity and without upsetting your venture investors:
- Restricted grants into a separate account. Keep the money apart, track it, and report that you used it for the charitable purpose. Some companies do this for many years before they set up anything more complex.
- Donor-advised funds. A donor who wants a tax deduction gives to a DAF, and the DAF makes a grant to your company for a charitable purpose. The DAF sponsor must be willing to do this and to follow up on how the money is spent. Many are not, so ask early.
- Program-related investments (PRIs) from foundations: loans or equity made for the foundation’s charitable purpose.
- Grants that convert. Newer instruments start as a grant and become debt or equity if the company reaches defined milestones.
Your existing investors should like pure grants: the money is non-dilutive, and it pays for safety work or low-margin work that the business would not fund. Be careful with instruments that must be repaid or redeemed. If the company does not do well, your other investors will see a funder that can pull money out ahead of them.
6. An independent public charity next to the company
Cost: a large part of a senior CFO’s time. When: prepare only, until you have that CFO.
In US law there are two main kinds of charity. A private foundation usually has one or a few funders and exists to give money away. It cannot do business with you: money flows out to unrelated grantees and nothing can flow back. A public charity that is independent of you can do business with the company. It can take on high-impact, low-margin work, license your technology, and pay the company for products and services. Flexport.org is a known example of a charity that works next to a company.
This is one of the most valuable and least used options. It is off your balance sheet, but it can add to your revenue, while a private foundation is pure cost. It is also too complex for a pre-revenue startup. Every flow between the two entities (money, services, IP) must be managed and documented, which can take most of a sophisticated CFO’s time. If you have that CFO and funders who want to give, consider it. Until then, use step 5.
7. The perpetual purpose trust
Cost: expensive to set up, hard to manage, hard to operate. When: prepare only.
The perpetual purpose trust (PPT) is arguably the best mission-lock mechanism anyone has built. It can serve a purpose with no human beneficiaries, forever. Patagonia’s trust exists to protect the company’s values and mission on behalf of the Earth. A PPT has stewardship trustees and a trust enforcer, and its governance is flexible: for example, you can say that whoever leads a named institution is always a trustee.
If you are a scrappy startup, do not set up a PPT now. Set up the company so that it can have one later. In practice that means step 2: a sentinel share class that a future trust can hold.
Traps
- A B Corp is not a legal form. It is a certification: you complete an assessment, pay a fee and sign a license agreement. It does not change your directors’ duties.
- Not every “benefit corporation” statute is equal. Some state benefit-corporation statutes are hard to invest in. For a venture-backed startup, the Delaware PBC is the default.
- Complexity has a cost. Every structure beyond step 5 needs people to run it. A lean team should count those hours before it starts. Restructuring a mature organization can take years.
A checklist for a pre-seed or seed company
If you want to make sure the company is not taken out of your hands when it succeeds, do this, in order:
- Write your values and link them to compensation, your public benefit and your reporting.
- Incorporate as (or convert to) a Delaware PBC with a short, measurable public benefit that describes what you do.
- Add a founder class vote over any change to the public benefit and any conversion out of PBC status. Make the shares transferable to a future trust or charity.
- If a founder owns key IP, license it in royalty-free with a mission lock. Decide who judges whether the company is still on mission.
- Add a springing public license on the company’s core IP, triggered by abandonment of the mission or bankruptcy.
- Write a one-page policy for restricted grants and open a separate account before the first foundation asks.
- Leave the sidecar charity and the perpetual purpose trust for later. Only make sure the charter does not block them.
A note from the investor side
In the last few years, the standard startup advice and the standard documents have converged. That makes fundraising faster. It also means the defaults now serve investors more than founders, because investors wrote most of them. Read Founders at Work and you will find very different companies, set up in very different ways. The best companies did not all use the same template.
The standard structure is one option among many. You can stay true to your values, raise venture capital and scale. We back founders who do this, and the Juniper Governance Lab exists to make it cheaper. If you are designing your structure now, write to us at team@juniper.xyz.