Skip to content
AGI Future Foundation PBC

Features · AGI Future Foundation

The Shield and the Windfall

Every technology company that ever promised to put safety first eventually met the same test: a quarter where doing the safe thing cost money, and a shareholder who asked why. The Foundation’s answer is to write the protection into the law before the money arrives — a shield for the board that chooses safety, and a windfall clause that decides, in advance, what happens if the bet pays off beyond imagining.

AI-generated illustration for The Shield and the Windfall
AI-generated illustration.

The race nobody wins by slowing down

Start with the incentive, because the whole architecture is built to survive it. A company building general intelligence faces a race: every rival is building, every month of delay is market share lost, and every safety check is a cost that a faster competitor may skip. Inside the company, the pressure concentrates on the board. Directors owe duties to shareholders, and shareholders, in the ordinary telling, want returns. A board that slows a launch to run one more evaluation, or turns down a lucrative deployment that fails a safety review, has made a decision that costs money in the short term. In a traditional company, that decision invites a lawsuit. Even the threat of one is usually enough. Safety loses not because anyone argues against it in the open, but because the legal structure quietly punishes the people who choose it.

That is the problem the Fiduciary Shield is designed to answer, and it is worth naming plainly before any of the mechanics arrive. The Shield does not make a company safe. It does not write a safety policy, run an evaluation, or stop a deployment. What it does is change who bears the legal risk of a safety-first decision. If a board can weigh public benefit alongside returns without being exposed for having done so, then safety stops being the choice that gets a director sued. The Shield is a piece of legal armor for a particular kind of decision, and the Foundation’s planning documents treat it as the first gear in a three-gear machine: the Shield protects the board, the Windfall Clause distributes the upside, and Open Global Investment spreads ownership widely enough that rivals have less reason to race the enterprise in the first place.

The stakes of getting this right are the reason the design is so careful about its own limits. A shield that promised too much would be worse than no shield at all, because it would let a board believe it was protected when it was not. The Foundation’s own paperwork caught exactly that overstatement in its pitch materials and corrected it, and this feature keeps the correction front and centre. Read on, and the first thing to understand is what the Shield is not.

What the Shield is, and what it is not

The Shield, in the corrected legal language of the Foundation’s documents, is expanded business-judgment protection for benefit-considering decisions. The mechanism is California public benefit corporation status. In draft form, the company’s Articles would carry a stated benefit purpose — advancing safe, equitable, and socially beneficial artificial intelligence and the global distribution of its benefits — which counsel must draft in final form. Under California Corporations Code sections 14620 and 14621, directors of a benefit corporation must consider that benefit purpose alongside shareholder interests. That duty is the heart of the Shield: a board that weighs safety and public benefit, documents its reasoning, and acts in good faith receives the expanded protection of the business-judgment rule when a profit-focused shareholder challenges the decision.

Concretely, that protection is what lets a board reject a buy-out that threatens safety on benefit grounds, rather than being compelled to sell to the highest bidder. It is what lets a board accept a slower, safer path without the decision itself becoming the lawsuit. And it is paired with an annual benefit report under section 14630, so the public-benefit side of the bargain is accounted for in public, year after year, rather than asserted once in a pitch deck and forgotten. The pitch materials call this a two-way check: accountable for both profit and public benefit, where a traditional company answers to a one-way ratchet for profit alone.

Now the correction, stated as plainly as the Foundation’s own pages state it. The Shield is not immunity. The pitch deck described the board as legally immune to shareholder lawsuits when sacrificing short-term profit for safety guardrails, and that claim overstates California law. A benefit corporation can still be sued. It can still be acquired. Its benefit purpose can still be amended by the required vote. The protection attaches to good-faith decision-making, not to outcomes, and it does not remove the duties themselves: a board that ignores the benefit purpose, or acts in bad faith, is not shielded by it. The honest phrase is that the Shield moves the litigation boundary — the line at which a shareholder can credibly sue over a safety-first decision shifts, because the law itself now asks the board to consider the benefit the decision protects. Moving a boundary is a real achievement. It is not a wall, and this feature never calls it one.

The Shield is also not a substitute for the rest of the architecture. It protects a board’s benefit-considering decisions. It does not, by itself, designate a series, seat a Steward Council, adopt a Windfall Clause in any operating agreement, screen a sovereign shareholder, or open a bank account. Each of those is a separate step, with its own documents and its own open items. The Foundation’s lawyer briefing memo counts the Shield at two evidenced items, seven open actions, and four counsel questions — a count drawn from the Foundation site rather than verified as law, but a fair picture of the shape of the work: most of the Shield is still open, and counsel must complete the open items before the Shield is relied on in any solicitation or offering document.

The Windfall Clause: deciding before the money exists

If the Shield is about protecting decisions made under pressure, the Windfall Clause is about removing the pressure from a decision that has not happened yet. Voluntary philanthropy has a well-known failure mode: a board that means to share profits later can be sued, outvoted, acquired, or simply worn down before the profits arrive. The moment of success is the worst moment to renegotiate generosity, because that is when the money is real, the shareholders are counting it, and every argument for keeping it sounds reasonable. The Windfall Clause answers that by making the sharing decision before the profits exist — hardwired into the governing documents, so sharing is the default rather than a decision re-argued at the moment of success.

As designed, the clause obligates all 33 W3 series — the profit-generating series in the April 2026 internal list — to share a portion of excess AGI profits toward international benefit sharing once a specific economic threshold is met. The threshold is expressed as a percentage of Global GDP, not a fixed dollar amount, because a dollar figure set today would be overtaken by the scale of the economy it measures, while a share of global output scales with the world it is drawn from. Below the threshold, profits follow the normal rules: standard economic returns to investors and retained funding for research and development. Above it, the excess becomes Windfall Profits, paid to a Windfall Administrator for distribution to four benefit buckets: Global Public Goods, Healthcare Access, Energy Infrastructure, and Technological Self-Determination, for projects advancing the UN Sustainable Development Goals. The buckets share equally unless the board and counsel set different shares, and no windfall distribution may go to a shareholder, director, officer, or series manager as a return on their shares, or be used to buy control, influence a government, or fund a political campaign.

Here is the part that makes this feature different from a sales pitch: no terms exist today. The threshold percentage is blank. The GDP measure — which source, which year, which vintage — is blank. The profit measure is blank. The scope — each series separately, the series in aggregate, or the parent plus the series — is blank. The Windfall Share itself is blank: the pitch calls it a massive portion, but gives no number, and none may be quoted as the share. The Administrator — the Foundation, a separate Windfall Fund, or a trust — is blank. The auditor, the dispute forum, the payment timing, and whether the clause survives a sale, merger, or change of control are all blank, for counsel to confirm in writing. A percentage of Global GDP with no percentage, no measure, and no scope is a direction of travel, not a trigger, and this feature will not describe it as triggered at any dollar figure.

Two guardrails are built into the design. First, the clause must be adopted before a series becomes profitable: one adopted after a series is already profitable, or set aside by a simple board majority when profits arrive, does not meet the standard. Second, it may be amended, suspended, or repealed only by a supermajority vote plus the Class C vote, with the exact percentage still blank for counsel to set. An amendment reducing the Windfall Share, raising the threshold, or removing a benefit bucket, adopted without that vote, would be void under the draft. The design allows amendment in a genuine emergency, but not by a simple majority at the moment the money arrives. And the clause must appear in each series’ designation documents, not only in a parent company’s articles: a parent-level clause alone binds no series, and the W3 series themselves are not verified as filed, so today the clause has no series to bind.

Three classes: splitting the money from the steering wheel

A shield protects a board, and a windfall clause decides what happens to excess profit. But neither answers the question of who owns the enterprise in the first place, and that question is where most grand designs quietly die. Ordinary ownership bundles two things together: a share of the economics and a share of the votes. Buy more of a company and you buy more say over what it does. That bundling is the reason broad ownership usually means broad control, and the reason a mission-driven company that sells equity usually ends up answering to the people who paid the most. The Foundation’s answer is to unbundle the two, in a three-class share structure that belongs to the for-profit benefit-corporation track only.

Class A carries one vote per share and full economic participation, and it is the broad-access class: the global public’s class, open worldwide subject to securities law in each jurisdiction. In the pitch diagram, Class A is almost entirely profit participation with only a small slice of voting power, and that is deliberate. Class A is how the world shares in the value AGI creates, up to the Windfall threshold, without handing safety decisions to the smallest voters. Broadly open does not mean open to anyone without checks: in the United States, Class A participation is subject to Regulation D, with no general solicitation unless Rule 506(c) plus investor verification is expressly chosen, and participation never carries a guarantee of liquidity, a secondary market, or a buyback. The phrase sometimes used for Class A — guaranteed participation in the economic upside — means participation in upside if there is upside, not a guaranteed return. Class A holders are told upfront what they are buying into: economics, not control.

Class B carries ten votes per share and is both economic and voting — roughly half economic participation and half voting power in the pitch diagram. It is reserved for restricted, aligned partners and sovereign entities meeting strict safety and geopolitical frameworks. That restriction is part of the standard, not a footnote to it: Class B is not open to any sovereign on demand, and a sovereign that does not meet the framework does not get Class B — and does not get Class B voting power by buying Class A instead. Any sovereign or rival participation needs, besides the offering documents, sanctions screening, export-control review, and national-security review where applicable, plus tax and securities advice per jurisdiction. The strategic logic is that a heavily invested counterpart has a vested interest in the enterprise’s secure, safe success. But that logic only works if admitted Class B participants are actually aligned and actually screened, which is why the framework, the qualified-holder definition, and the screening criteria are all blanks for counsel to set before any Class B solicitation. Transfer rules are likewise for counsel: transfer only to a qualified holder, or conversion to Class A on transfer to a non-qualified holder.

Class C carries 1,000 votes per share, is held exclusively by the Foundation, and has no economic participation at all — control only. The voting math is the mechanism. At 1,000 votes per share, even a small number of Class C shares dominates Class A at one vote and Class B at ten votes on any ordinary vote. That is how the Foundation acts as the ultimate ethical backstop: shielding the profit-generating series from hostile market pressures, keeping operations consistent with AI safety and equitable distribution, and preventing ruthless profit maximization from overriding the mission. The Foundation’s control over the W3 series runs through Class C control at the parent level, mission and voting agreements with each series, and each series’ Steward Council mission gate — not through owning each series’ economics. Control is not ownership, and the distinction matters: the Foundation does not own the series’ profits, and its control does not entitle it to them.

Two limits stay visible on Class C, and both are worth stating because they are the difference between a safeguard and a throne. First, founders must not be issued Class C: it is Foundation control for the mission, not founder control, and founder placement is a counsel decision, with the offering documents required to state in one paragraph who holds each class and why. Second, Class C is non-transferable except to a successor Foundation board or trustee as the Articles provide, and counsel is to set a conversion or cancellation trigger so control does not become a perpetual unaccountable block. A control class with no expiry and no accountability would reproduce the very concentration of power the structure exists to prevent, only under a different flag.

The disclosure duty follows the decoupling, and the Foundation’s documents are explicit about it. The Windfall Clause must be presented as a cost, not a benefit, and Foundation control as a governance fact, not a guarantee. Class A and Class B investors must be told, plainly, that they own economics and do not control safety decisions, and that Class C’s control is not a promise that the clause will never change. A structure that hides its own concentration of power behind elegant language about decoupling would be a deception with better branding. The documents refuse that shortcut, and this feature follows them.

Open Global Investment: inviting the rivals in

The share classes are the machinery. Open Global Investment is the strategy that machinery serves, and it starts from a blunt observation about how nations behave. AGI kept as a closely guarded national asset gives every capable party the same incentive: race to build it first, keep others out, and accept safety risks rather than fall behind. Unilateral development incentivizes espionage, sabotage, and rushed, unsafe deployment, because each side fears exclusion and dependency. A program benefiting only one country concentrates the transition’s wealth in one place and leaves everyone else with a reason to undermine it. That is not a moral judgment about any particular country. It is a description of what the incentives do to everyone who faces them, and the Foundation’s planning documents treat it as the central strategic problem of the AGI transition.

Open Global Investment is the proposed answer, and its logic is deliberately counterintuitive. Instead of excluding rivals, invite them into the economic upside. Instead of concentrating ownership, spread it widely. A party with a real stake in the safe success of the enterprise has less reason to race recklessly against it or try to seize it, because undermining the enterprise would mean undermining its own stake. That is the co-opting rivals instead of racing them idea at the centre of the model, and it is the reason Class A is designed for global public access while Class B gates sovereign participation behind a framework rather than a door. Everyone can own the future, but the future’s safety cannot be subjected to a hostile takeover — that is the pitch framing, and the share structure is what makes it more than a slogan.

The model is unproven, with no terms, no participants, and no track record today, and the honesty rules of this feature require saying so in the same breath as the idea itself. The participation instrument is blank: Class A shares, profit participation interests, a revenue share, or something else, with counsel to choose one and provide a securities analysis for it. Eligibility is blank: who may participate, accredited only, international participants, rival or sovereign participants, with securities, sanctions, tax, and anti-money-laundering advice per category. Price and valuation are blank, set at the participation offering, not today. Governance rights are blank, with a default of none. Transfer and liquidity are blank, with no liquidity, secondary market, or buyback promised. Reporting is blank, as counsel advises. It is not a SAFE, not a priced round, and not a donation — a separate program needing its own documents, not to open until Tranche 2 terms, the annual benefit report, and counsel’s signed Tranche 3 terms are all in place.

The proposed for-profit PBC raise is organised in three tranches, and every figure here is a suggested planning framework, not a price, not a term, not a promise. Tranche 1 is a post-money SAFE: a suggested round of $500,000 to $1,500,000, a suggested post-money cap of $8,000,000 to $15,000,000, and a discount of 20% or none — counsel and the board must choose one discount, not both. The suggested minimum is $25,000 per investor. Investors would be accredited only, under Regulation D, with a disclosure document and a Form D within 15 days of first sale, and SAFEs would convert on a qualifying priced financing into Class A shares. Tranche 2 is a priced equity round, priced preferred stock or Class A common as counsel advises, with its price set at Tranche 2, not today, by the board, a lead investor, and counsel, from actual traction at that time. Tranche 3 is the Open Global Investment participation itself, with the Windfall Clause. Nothing has been issued under any tranche. No cap, discount, or round size is set today.

Two alternative issuers exist for any future raise, and only one can be used. The for-profit benefit-corporation track described here and the AGI Corp SAFE in the other planning documents are alternative issuers, not both at once for the same round. The AGI Corp SAFE was built around AGI Corp, the intended Wyoming corporation, as issuer, with the Foundation as nonprofit mission steward. The PBC structure builds an alternative issuer — a for-profit benefit corporation — with dual-class shares, a benefit purpose, Open Global Investment, and Windfall mechanics. Running an AGI Corp SAFE and a PBC SAFE for the same money, to the same investors, on different terms, would leave investors unsure which entity they were buying. Counsel must choose one issuer and one track in writing before any solicitation, so investors know which entity they are buying, and no choice has been verified.

The fork: two things called PBC

There are two different things called AGI Future Foundation PBC in the family’s materials, and they are not the same entity type. This is the single most dangerous confusion in the pack if a reader mixes the two stories, and it is the reason this feature spends a full section on it rather than a footnote. As filed and verified, the Foundation is a California Nonprofit Public Benefit Corporation, Entity No. B20260196664, Articles filed 2026-04-24, with Secretary of State approval 2026-04-30, principal and mailing address 995 Market St AGI Lab -10, San Francisco, CA 94103, agent Mike Rice, and an EIN assigned 2026-05-07. Its purpose carries 501(c)(3) language, with irrevocable dedication and no private inurement. A nonprofit with this filing cannot sell equity, SAFEs, profit shares, or tokens, and cannot distribute profits.

As labelled on the live website — including on the site’s home page, which calls the Foundation a California Public Benefit Corporation and describes the Shield as a for-profit benefit-corporation defense — PBC is used in the for-profit sense: a benefit corporation with dual-class equity, Class A profit participation, and Open Global Investment. The corporate raise advice follows the filed status, because a court, the Attorney General, the IRS, and an investor’s counsel will follow the filing, not the website label. If a for-profit benefit corporation is intended, that is a different corporation to form, or an amendment or conversion that counsel must advise on. A nonprofit cannot simply become a for-profit and keep its donations.

The fork is unresolved. Until counsel chooses one track in writing, no solicitation should use the for-profit figures, and the Shield as described on this page must be understood as the for-profit PBC track’s design, not as a protection already held by the filed nonprofit. The raise advice is explicit that the website’s Class A, dual-class, and Open Global Investment copy describes a for-profit equity structure that does not apply to the filed nonprofit, and that using it to solicit money for the Foundation risks both a securities violation and a charitable-solicitation violation. For the nonprofit track itself, the rule is simple: the filed Foundation raises by donation and grant only. It cannot sell shares, SAFEs, convertible notes, tokens, profit participation, Class A shares, or Open Global Investment interests, and it cannot promise a donor a return, a windfall share, or a share of series profits. No SAFE by the filed nonprofit Foundation is permitted.

Two facts stay attached to this page. First, no IRS 501(c)(3) determination letter is in the files for the Foundation — the EIN notice lists Form 1120 — so the Foundation must never be presented as IRS tax-exempt, and no donation may be promised as tax-deductible. Second, the Foundation’s Statement of Information, initially due around 2026-07-23, and Attorney General Registry registration are not evidenced in the files, and the review of 2026-11-03 records them as apparently past and unverified — status to be checked, not assumed. The Shield does not cure any of that. A reader who wants the filed facts and nothing else should start with the governance pages, which state the fork in full, and treat this feature as the story of a design, not the record of a company.

There is also a structural tension inside the intended design that belongs in the same passage as the design itself. The intended structure places AGI Corp, a Wyoming corporation, as the commercial head of the family, with the Foundation beneath it as steward. A California nonprofit cannot be controlled by a for-profit head the way a subsidiary is controlled. The Foundation must keep an independent board, independent decisions, and the right to refuse a project that fails its mission standards, even if AGI Corp wants it. Until counsel designs that control mechanism, the honest description is that the Foundation is a steward, not a captive, and that control runs only through written mission agreements between equals. The Shield, the Windfall Clause, and the share classes are all designs for a corporate holding whose foundations — which entity issues what, and who controls whom — are still being drawn.

What remains, and what is not true yet

The governance described in this feature is a design to be built, not a system already in force, and the remaining work is the honest ending of the story rather than a postscript to it. The Articles’ benefit purpose must be drafted by counsel in final form, and the sections 14620 to 14621 duty analysis and the section 14630 annual benefit report confirmed in writing. The Shield’s seven open actions must be completed and its four counsel questions answered. One PBC track must be chosen in writing — a new for-profit benefit corporation, a conversion, or AGI Corp as issuer — before any solicitation, and the site copy reconciled with the choice. The Foundation’s Secretary of State good standing, Statement of Information, Attorney General registration, and IRS posture must be confirmed before any donation round or bank application. The board that would rely on the Shield must be seated, with signed bylaws and consents, a third director, and the pending co-founder consent resolved in writing. The pitch’s legally immune claim must be replaced with the corrected expanded business-judgment protection language everywhere it still appears. And counsel must confirm what insurance each entity should carry before fundraising, rather than treating the Shield as a substitute for coverage.

Nothing has been issued. No shares, no SAFE, no convertible note, no token, and no profit-participation interest has been issued by the Foundation, by AGI Corp, or by any series. No offering is open, and nothing on this page is an offer to sell securities. No SAFE or equity is offered by the filed nonprofit. Formation is unverified beyond the two nonprofits: only the Foundation and the Fellowship have verified Secretary of State filings in the files reviewed, AGI Corp has an EIN notice only, and W3 LLC, the Delaware cohort master, any for-profit benefit corporation, and any SmartCiti.X corporation are intended, not verified as filed. No site, union, customer, or revenue is claimed here: no site has been purchased, leased, or placed under contract, no union or pilot customer has signed, and no revenue is evidenced in the files reviewed. Suggested valuations, raise sizes, tranche splits, and use-of-proceeds figures in the planning files are illustrative planning figures only — not prices, not appraisals, not guarantees, and never to be summed into a family total.

That is a long list of not-yet, and it is the point. The Shield and the Windfall are ideas about how to govern power before power arrives, and their value is in the thinking they force: what would have to be true for a board to choose safety under pressure, for excess profit to flow to humanity by default rather than by generosity, for rivals to own a stake in a safe outcome rather than a reason to wreck it. Those are questions worth asking while the answers are still blank, because once the money exists, the questions get much harder to ask. The Foundation’s documents ask them early, write down the blanks honestly, and refuse to fill them with invented numbers. That discipline — the willingness to publish a design with its holes showing — is the most convincing part of the design. A reader who finishes this feature should know exactly what the Shield protects, exactly what the Windfall promises, and exactly how much of either exists today: the ideas, the drafts, and the open questions. The rest is work still to be done, by humans, with counsel, in writing.

What this is / What it isn’t

What this is

  • A draft governance design: expanded business-judgment protection for benefit-considering decisions, a pre-commitment to share excess profits, and three share classes that separate economics from control.
  • A story about incentives: why a board needs legal room to choose safety, and why sharing must be decided before profits arrive.
  • A public record of what is blank: threshold, share, administrator, issuer, and every operative term, left for the board and counsel to set.

What it isn’t

  • Not immunity: the Shield is expanded business-judgment protection, and a benefit corporation can still be sued, acquired, or amended by the required vote.
  • Not an offering: no shares, SAFE, or participation interest has been issued, and nothing here is a price, a term, or a promise of return.
  • Not in force: the for-profit PBC track is unresolved, the filed entity is a nonprofit, and the W3 series the Windfall Clause is designed to bind are intended, not formed.

Where to see it

The governance pages carry the full detail behind this feature: Fiduciary Shield, Windfall Clause, Share Structure, and Open Global Investment, with the whole architecture on the Governance hub. The Shield’s repository is public on GitHub: FiduciaryCorporateShield, described there as an agent based fiduciary corporate shield. Its README is two lines, title only, and the companion shield-site is a stub: no substantive code was reviewed behind the Shield. The Shield lives in the legal and governance documents, not in code, and this feature presents it that way. The PBC in the repository’s branding copy is site copy, not the filed status: the filed entity is the California nonprofit described in the footer below.

Nothing on this page is legal, tax, or financial advice. The Shield and the Windfall Clause are draft designs for a for-profit benefit-corporation track that counsel has not yet chosen; the filed entity is a California nonprofit public benefit corporation, and no IRS 501(c)(3) determination letter is in the files.

← Prev: Worlds on Real Maps   Next: The Watchers →