Risks and Mitigations: An Honest Investor FAQ

**SEO meta description:** AGI Future Foundation PBC answers the hardest investor questions about governance, safety, legal structure, and commercial viability — directly, without spin.

Introduction: Why We Lead With Risk

The standard investor presentation leads with opportunity and buries risk in a footnotes section labeled "Risk Factors." The Foundation takes the opposite approach in this document. Sophisticated investors who are seriously evaluating a novel governance structure for a high-stakes technology deserve a frank account of what can go wrong and what we have done about it — before they read about the upside.

This FAQ addresses the questions that diligence investors most commonly raise, organized by the category of risk they represent. Where we have mitigations in place, we describe them precisely. Where the risk is not fully mitigated, we say so. Where a risk requires independent expert review to assess, we say that too.

Q: Is the three-tier structure legally tested?

**A:** No, not in the specific combination the Foundation uses. California Public Benefit Corporation law is well-established, as is Wyoming's Series LLC statute and DAO LLC supplement. Dual-class share structures are common in Delaware corporations and increasingly used in California. But the combination of a California PBC apex, a Wyoming Series LLC commercial engine with 33 DAO subsidiaries, and a governance architecture designed specifically for AGI safety governance has not been tested in litigation or in regulatory proceedings.

This is a frontier legal structure. It is designed by counsel with relevant expertise in each component, but it has not been validated by courts or regulators. Investors should obtain independent legal opinion on: (a) the enforceability of the series liability isolation across jurisdictions where the series operates; (b) the legal status of DAO membership interests under U.S. securities law; (c) the scope of PBC fiduciary protection for specific types of safety-based decisions; and (d) the Windfall Clause's enforceability as a governance mechanism.

**Mitigation:** The Foundation employs counsel with expertise in each component area and maintains a legal risk register that tracks open questions and the state of legal guidance on each. The Governance Tracker logs legal review milestones. This reduces but does not eliminate legal uncertainty.

**Residual risk level:** Medium-high. The legal architecture is sound in design; enforceability in specific contested scenarios is not fully known.

Q: Can the Fiduciary Shield actually prevent shareholders from forcing unsafe decisions?

**A:** The Fiduciary Shield significantly constrains the ability of shareholders to force safety-compromising decisions through legal mechanisms. The California PBC statute modifies the fiduciary duty framework; the dual-class structure limits voting control; and the business-judgment rule provides a defense for mission-based decisions documented through the Governance Tracker.

However, "significantly constrains" is not "fully prevents." The following scenarios represent residual risks:

- **Litigation risk:** A large shareholder could challenge a Foundation board decision in California court, arguing that the board's invocation of PBC mission authority was pretextual or that the decision exceeded the scope of the PBC's stated purpose. Even if the Foundation ultimately prevails, such litigation is expensive, distracting, and reputationally disruptive. - **Social and market pressure:** Legal protections do not prevent informal pressure — public campaigns, media narratives, partner relationships, or key employee defections — designed to force governance changes outside of formal legal channels. - **Board capture:** The dual-class structure protects against external shareholder pressure but does not protect against the risk that Class B shareholders themselves become misaligned with the mission over time. Board governance — director selection, evaluation, and succession — is the appropriate check on this risk, and it requires ongoing attention.

**Mitigation:** The Governance Tracker's public-facing record, the Foundation's transparency commitments, and the procedural requirements that accompany any board governance change create friction against all three of these scenarios. Independent governance audits are planned to assess board composition and independence over time.

**Residual risk level:** Medium. Legal structure is sound; social and internal governance risks require ongoing management.

Q: What happens to the Windfall Clause if the Foundation is acquired or dissolved?

**A:** The Windfall Clause is designed as a durable contractual commitment, not merely a board policy that can be reversed by a board vote. Its specific enforceability in an acquisition or dissolution scenario depends on how it is incorporated into the Foundation's charter documents, the relevant intercompany agreements, and California PBC dissolution law.

California PBC law requires that on dissolution, assets must be distributed to another public benefit corporation, to the state, or to another entity with a compatible charitable purpose. This provides some baseline protection against a scenario in which commercial acquirors obtain the Foundation's assets and simply redirect them to private purposes. But an acquisition structured as a merger, or a conversion from PBC to a different corporate form, could potentially affect the Clause's enforceability depending on how it is legally embedded.

**Mitigation:** The Foundation's legal counsel is tasked with embedding the Windfall Clause in a manner that is as dissolution-resistant as possible under California law, including provisions that require court approval for charter amendments that affect the public-benefit purpose. Independent legal review of the Clause's specific drafting is required to assess how durable it is in non-standard exit scenarios.

**Residual risk level:** Medium. Baseline statutory protections exist; edge-case enforceability requires legal review.

Q: Does passing the alignment evaluation gate mean a model is safe to deploy?

**A:** No. Passing the gate means the model did not exhibit the specific behavioral failure modes tested under the specific evaluation conditions in the harness. This is meaningful evidence, but it is not proof of safety in deployment.

The gap between evaluation and deployment is real and unavoidable with current alignment science. A model that behaves well on deception probes in a controlled evaluation may behave differently when deployed in a commercial context with millions of users, novel prompting patterns, adversarial inputs, or emergent multi-agent interactions that the probes did not anticipate. Spec-gaming — behaving differently during evaluation than during deployment — is one of the specific failure modes the harness tests for, but testing for spec-gaming is not the same as eliminating it.

**Mitigation:** The Foundation's approach to this limitation includes: (a) post-deployment monitoring requirements attached to all conditional and standard gate approvals; (b) rapid response protocols for behavioral anomalies identified in deployment; (c) incremental deployment scopes (limited user populations, restricted use cases) as standard practice for initial releases; and (d) ongoing expansion of the probe library as new failure modes are identified in the field and in the alignment research literature.

**Residual risk level:** High (inherent to current alignment science). The gate is a meaningful risk-reduction mechanism; it is not a safety guarantee.

Q: What if the independent LLM judge is itself misaligned?

**A:** This is a genuine risk and one that the Foundation takes seriously. The independent LLM judge is used to evaluate candidate models at scale — if the judge is itself deceptive, miscalibrated, or susceptible to spec-gaming, it could produce passing scores for models that should fail.

The Foundation's current mitigations: - The judge is sourced from a provider independent of AGI Corp's development pipeline, reducing the risk of correlated failure modes. - The judge's scoring is reviewed by fiduciary oversight, which can reject or adjust judge outputs that appear anomalous. - The judge is itself subject to periodic evaluation against human expert scoring on a sample of probe responses, to verify that its scoring remains calibrated. - The judge model is rotated periodically to prevent optimization by candidate models against a known judge architecture.

**Residual risk:** These mitigations reduce the risk of judge failure but do not eliminate it. An adversarially capable model that knows the general characteristics of the judge category could potentially optimize for judge approval while failing in deployment. This is a frontier research problem in alignment science, not a solved one.

**Residual risk level:** Medium (with active mitigations). Will remain an area of ongoing development.

Q: What prevents the probe set from being gamed over time by the development team?

**A:** Several structural constraints make this difficult. First, the Coverage Veto gives fiduciary oversight the right to reject probe sets that appear insufficient — including sets that are technically comprehensive but practically hollow because they test only the specific behaviors the development team has already optimized against. Second, the probe set is not public; its specific scenarios are proprietary, reducing the risk of direct optimization during training. Third, the Change Control rule voids any run in which probe changes are made after pre-registration, preventing last-minute probe softening.

The more subtle risk is that over time, a development team that submits probe sets repeatedly learns what coverage is considered acceptable and designs future probe sets that satisfy coverage review without being genuinely challenging. This is a social dynamics risk, not a structural one, and it is addressed through: (a) external participation in probe design from the alignment research community; (b) periodic external audits of probe adequacy by parties with no relationship to the Foundation's development team; and (c) the Coverage Veto as an ongoing check.

**Residual risk level:** Medium-low in the short term; requires active management over longer periods.

Q: Can the Foundation generate returns that justify institutional investment while maintaining safety constraints?

**A:** This is the central commercial question, and the honest answer is: we believe so, but the proof is in the execution and has not yet been fully demonstrated.

The commercial thesis is that trustworthy AI capability — capability that enterprise customers can deploy with confidence that it will not produce liability exposure, reputational harm, or alignment failures — commands a premium in the market for AI services. If this thesis is correct, then the cost of the Foundation's safety infrastructure is offset by the price premium it enables and the market segments it opens (regulated industries, high-stakes applications, government procurement) that less-governed competitors cannot credibly serve.

The risk is that the market does not actually pay a meaningful premium for safety governance, or that competitors catch up on safety credibility without matching the Foundation's governance costs, compressing margins. If safety becomes expected table stakes rather than a differentiator, the Foundation's governance investment becomes a cost without a corresponding revenue advantage.

**Mitigation:** The Foundation's commercial strategy emphasizes market segments where the cost of AI failure is high — healthcare, financial services, legal, and critical infrastructure — because these are the segments where safety governance commands the largest premium and where regulatory requirements make governance documentation (like the Governance Tracker) a procurement prerequisite rather than a differentiator. This concentrates commercial exposure in segments where the thesis is most likely to hold.

**Residual risk level:** Medium. Thesis is credible; market validation is ongoing.

Q: What is the risk that a 66-W3 series fails and creates contagion?

**A:** The Wyoming Series LLC structure is designed to provide liability isolation between series — a failure in one series should not attach to the assets of other series or the master LLC. This isolation is a core design principle of the commercial structure.

The legal risk is that series isolation is a newer and less-tested doctrine than standard LLC liability isolation. Courts in other states — or courts applying other states' laws to Wyoming series assets — may not recognize the isolation in specific fact patterns. The practical risk is that reputational damage from a high-profile series failure could affect the Foundation's ability to attract commercial partners or investors for other series, even if the legal isolation holds.

**Mitigation:** Ongoing legal monitoring of series LLC case law across jurisdictions; operating agreement structures designed to maximize isolation under multiple legal frameworks; and the Foundation's governance transparency, which means the Governance Tracker provides a clear record distinguishing the failed series' governance history from other series that operated under appropriate controls.

**Residual risk level:** Medium. Legal isolation is sound in design; full cross-jurisdictional enforceability requires legal review.

Q: What is the exit path for investors?

**A:** The OGI model's dual-class structure, the Windfall Clause, and the PBC status together constrain the exit options that would be familiar to investors in standard venture-backed companies. Specifically:

- **IPO:** A California PBC can be publicly traded, but the dual-class structure and PBC governance constraints would need to be disclosed to and accepted by public market investors. This is feasible but requires careful structuring and market-timing judgment. - **Acquisition:** An acquisition of the Foundation would need to be structured in a manner consistent with PBC dissolution requirements, meaning the acquiror would need to be a compatible public-benefit entity, or the acquisition would need court approval. A straightforward commercial acquisition by a profit-maximizing buyer is structurally difficult. - **Secondary liquidity:** Secondary sales of Class A economic interests may be available through structured transactions, depending on the specific terms of the investment agreements. - **Windfall distributions:** If the commercial engine generates returns that trigger the Windfall Clause's redistribution mechanism, investors in Class A shares receive distributions according to the Clause's terms before redistribution to broader beneficiaries.

**Investors should clearly understand:** The Foundation's structure is specifically designed to prevent the exit path in which a financially successful AI capability company is acquired by a large technology company for its capabilities, without its governance constraints, and its safety commitments are abandoned post-acquisition. This design is a feature, not a bug, from the mission perspective. From the investor perspective, it means that exit flexibility is less than in a standard venture investment.

**Residual risk level:** High relative to standard venture structures (intentionally). Investors for whom exit flexibility is a primary criteria should assess this carefully before committing.

Q: What is the regulatory exposure if DAO membership interests are securities?

**A:** This is a live legal question that the Foundation's counsel is actively managing. Under U.S. securities law, an interest in a DAO that is offered to investors with an expectation of profit derived from the managerial efforts of others may satisfy the Howey test for an investment contract and thus constitute a security. If DAO membership interests in the 66-W3 series are securities, they must either be registered under the Securities Act of 1933 or fall within an exemption.

The Foundation's legal strategy for each series includes analysis of whether that series' membership interests constitute securities and, if so, how to structure the offering under an available exemption (Regulation D, Regulation S, or other applicable exemptions depending on the investor class and geography). This analysis is series-specific and fact-dependent.

**Residual risk:** Regulatory guidance on DAO tokens and interests is evolving and is not uniform across jurisdictions. The risk of a regulatory finding that a series' interests were not properly registered or exempted is real, though the Foundation's legal strategy is specifically designed to minimize it.

**Residual risk level:** Medium. Active legal management is in place; regulatory uncertainty is inherent in the space.

Q: What if a model the Foundation releases causes harm despite passing the gate?

**A:** This is the most significant reputational and potentially legal risk the Foundation faces. If a model that passed the alignment evaluation gate causes documented harm in deployment — even if the harm was not predictable from the gate's probe domains — the Foundation's governance credibility will be tested severely.

The Foundation's response framework for this scenario includes: - **Post-deployment monitoring:** All released models are subject to ongoing behavioral monitoring; anomalies trigger a review process that can result in model modification, restricted access, or withdrawal. - **Rapid response protocol:** If harm is identified, the protocol prioritizes harm reduction first (access restriction, deployment pause) and governance review second (Tracker logging, root cause analysis, threshold review). - **Transparent communication:** The Foundation's governance credibility depends on not minimizing or concealing harm incidents. The Governance Tracker's record of the gate process provides the factual foundation for an honest post-incident account. - **Remediation requirements:** The gate's conditional-pass and fail pathways exist precisely because the Foundation expects that some model versions will require remediation before release. The existence of these pathways reduces the pressure to force a marginal model through the gate.

**Residual risk level:** High (inherent to the frontier technology). Active mitigation is in place; this risk cannot be eliminated.

Conclusion: Why We Tell You This

The Foundation tells investors about these risks not because we are required to (though some disclosure requirements may apply) but because we believe that an investment relationship built on accurate risk understanding is more durable and more productive than one built on a sales presentation.

Investors who understand these risks and choose to invest are investors who have made a considered decision that the Foundation's approach — its governance structure, its safety commitments, its commercial strategy — represents a credible bet on a difficult problem. Those are the investors we want. Investors who are surprised by these risks after committing are investors who have been poorly served, and poorly served investors eventually become former investors and litigants.

The Foundation's governance transparency is not a disclosure strategy. It is an operating philosophy. The risks described in this document are real. The mitigations are genuine. The residual risks are honestly labeled. That is the standard we hold ourselves to, and it is the standard we invite you to hold us to as well.