How do courts decide whether AI-built wealth is marital or separate property, and how is AI company equity valued in divorce?

The Direct Answer: AI Wealth Follows Traditional Property Rules, With New Valuation Problems

The classification of AI-built wealth in divorce follows the same legal framework that courts have applied to startups, hedge funds, and private equity for decades: property acquired during the marriage is generally marital (community) property subject to division, while property acquired before marriage, by gift, or by inheritance remains separate property. What makes the current wave of AI divorces different is not the classification doctrine but the valuation problem. When a founder builds an AI company during a marriage — as happened in the SK Group chairman's divorce in South Korea, where a ruling on an asset division exceeding $1 billion was handed down after years of litigation — the court must assign a dollar figure to illiquid equity, unvested options, and speculative future revenue from models that may be obsolete within eighteen months.

Also worth reading: How Are Courts Sanctioning Lawyers for AI Citation Mistakes in 2026? · Verifying AI Citations for Courts: What Lawyers Must Do in 2026? · Can courts sanction a party for AI-related spoliation in eDiscovery, and what do lawyers actually need to do in 2026?

Courts in equitable distribution states (the large majority of US states) divide marital property fairly but not necessarily equally, while community property states like California and Texas start from a presumption of a 50/50 split. In either system, the timing question dominates: was the AI company founded before the marriage, during it, or did its value appreciate substantially during it? A founder who incorporated an AI lab three years before marrying but whose valuation multiplied fifty-fold during the marriage will face claims against the appreciation, even though the underlying shares remain separate property. This distinction between the asset itself and its marital growth is where most AI-founder divorces are actually fought.

The second layer of complexity is liquidity. Unlike a brokerage account, equity in a private AI company cannot simply be sold and split. Courts must either offset the value against other assets, structure deferred payments, award the non-founder spouse a percentage of sale proceeds, or use constructive trust mechanisms. Each approach carries tax consequences and risk allocations that can swing the effective recovery by tens of percent. The SK chairman's case illustrates the scale: when the asset pool exceeds $1 billion and most of it is concentrated in a single operating company, the difference between two competing valuations can exceed the entire net worth of a typical litigant.

Finally, AI wealth raises evidentiary questions that traditional discovery tools struggle with. Code repositories, model training logs, GPU purchase records, and cloud billing histories are now legitimate sources for tracing when value was created. Firms practicing in this space increasingly use AI-powered eDiscovery platforms to process terabytes of technical records, because manual review of engineering documentation at this scale is neither affordable nor reliable.

Why Classification Is Harder With AI Companies Than Traditional Businesses

A restaurant or retail chain built during a marriage has a fairly legible value history: revenue, locations, payroll records. An AI company's value often exists before revenue does. Foundation model labs have been valued at billions of dollars while generating modest or negative cash flow, meaning the 'marital effort' standard — which asks whether the increase in value resulted from labor, skill, or contributions during the marriage — must operate on metrics like model benchmarks, talent retention, and compute capacity rather than earnings.

Consider the Two Sigma divorce trial coverage that examined when AI-built wealth becomes marital property: quant funds and AI research firms generate returns through proprietary algorithms whose value is embedded in human capital and trade secrets. If one spouse spent the marriage building trading models or training infrastructure, the non-founder spouse's counsel will argue that every dollar of appreciation reflects marital labor. The founder's counterargument typically invokes active versus passive appreciation doctrine: passive growth (market forces, industry-wide AI hype cycles) stays separate, while active growth attributable to the founder's post-separation efforts may be carved out. In practice, separating an AI company's 2023–2026 gains into 'market beta' from the sector-wide boom versus 'alpha' from individual effort is nearly impossible with precision, and dueling experts routinely produce valuations differing by 40–70%.

Vesting schedules add another trap. Options and restricted stock units granted during the marriage but vesting after separation are treated differently across jurisdictions. Some states apply time-rule formulas apportioning value between marital and post-marital periods; others classify unvested grants entirely by grant date. For an AI executive whose compensation is 80%+ equity, this single doctrinal choice can move millions of dollars. Deferred compensation adds yet another wrinkle — Law.com has covered how unfunded retirement plans at professional firms are handled on divorce, and similar unfunded promises (phantom equity, profit interests) are common in AI startups where cash compensation is kept low to extend runway.

There is also a control dimension documented in National Law Review commentary on founders' divorces: family courts can inadvertently force dilution of founder voting power if they order transfers of common stock. Sophisticated settlements therefore separate economic value from governance rights, giving the non-founder spouse a financial interest while the founder retains board control — critical when the company's valuation depends on key-person continuity that investors scrutinize.

Practical Steps: How to Actually Value AI Equity in a Divorce

The first step is assembling a complete asset map, which in AI cases means going beyond statements. Counsel should collect cap tables from every funding round, 409A valuation reports, SAFE and convertible note terms, option grant agreements with vesting ledgers, RSU confirmations, phantom equity or profit-interest agreements, and any side letters granting liquidation preferences. Cloud computing spend and GPU procurement contracts serve as proxies for infrastructure investment timing, helping experts date when value creation occurred relative to the marriage.

Second, engage a qualified business appraiser with technology-sector experience early — ideally before temporary support hearings, since interim rulings can lock in positions. Standard approaches include the income method (discounted cash flow on projected model licensing or subscription revenue), the market method (comparable transactions among AI companies), and for pre-revenue labs, cost-based or recent-financing-round methods adjusted for lack of marketability discounts, which commonly run 20–35% for private company interests. Expect each side's expert to fight over the discount rate: a 2% change in the weighted average cost of capital applied to a high-growth AI projection can alter value by 30% or more.

Third, use modern eDiscovery tooling to trace value creation. AI-assisted document review platforms can process Slack channels, commit histories, email, and investor updates to build a timeline showing which milestones were achieved during the marriage. This matters both for the active/passive appreciation argument and for detecting hidden assets — undisclosed token holdings, offshore entities holding IP, or side agreements with acquirers. Discovery obligations in most jurisdictions require disclosure of all interests, including witness statements, valuations, and account statements that would otherwise surface at hearing.

Fourth, negotiate structure before litigating number. Common structures include: an equalization payment funded over 5–10 years with interest; a percentage interest in net proceeds upon a liquidity event (sale, IPO, or tender); an offset using liquid assets and real estate; or a royalty-style arrangement tied to future distributions. Each shifts risk. A fixed equalization payment protects the non-founder spouse if the AI sector corrects; a proceeds percentage protects the founder if the company exits above the trial valuation. Given that private AI valuations swung dramatically between 2021 and 2024 funding environments, risk allocation deserves as much negotiation attention as the headline number.

Fifth, address tax gross-ups. Dividing pre-tax equity without adjusting for capital gains exposure systematically overpays the recipient spouse. A share of stock with a $10 million paper value and a near-zero basis carries roughly $2 million less in after-tax value than $10 million in cash, depending on holding periods and applicable rates.

Comparison: Valuation Methods and Settlement Structures Side by Side

On the structural side, compare the dominant settlement vehicles. A fixed equalization note gives certainty and closes the case, but exposes the recipient to collection risk if the founder's liquidity depends on a delayed exit. A contingent percentage aligns both spouses' incentives around maximizing company value but keeps them financially tethered for years, which some judges view as prolonging entanglement contrary to the clean-break policy many states prefer. Offsets using houses and retirement accounts preserve the founder's equity intact but require the recipient to accept concentration in other assets. There is no universally superior option; the right choice depends on the company's stage, the couple's other assets, and each party's risk tolerance.

Mediation deserves mention here as an alternative to trial. The practice dates back to Ancient Greece, where the non-marital mediator was known as the proxenetas, and it remains the fastest route to resolution in high-net-worth cases. Mediation allows parties to adopt creative structures — staged buyouts, IP licenses between ex-spouses, confidentiality provisions protecting trade secrets — that judges rarely order. It also keeps technical details like model architectures and customer lists out of public filings, a real concern given press attention to cases like the SK chairman's billion-dollar-plus ruling.

Common Mistakes That Cost Founders and Spouses Millions

The most expensive mistake is treating the last preferred financing round as gospel. Strategic investors sometimes price rounds above what arm's-length buyers would pay, particularly during AI hype cycles, and courts know this. Conversely, founders who argue their company is worth far less than the round price while simultaneously telling investors and press that the technology is transformative destroy their own credibility. Consistency across investor decks, press interviews, and court testimony is essential; opposing counsel will subpoena all three.

The second mistake is ignoring vesting and acceleration mechanics until late in negotiations. Change-of-control clauses, double-trigger accelerations, and repurchase rights materially affect what equity is actually worth to each spouse. A settlement built on gross option values without modeling the exercise cost, expiration dates, and blackout windows produces numbers no one can realize.

Third, spouses frequently overlook intangible assets specific to AI ventures: trained model weights, proprietary datasets, patent applications, and personal brand value. Whether model weights developed partly with marital resources constitute divisible property is unsettled, but leaving them undiscovered guarantees they escape division. Similarly, deferred or unfunded compensation arrangements — the subject of dedicated Law.com analysis regarding law firm retirement plans — exist in analog form throughout tech companies and are routinely missed in disclosure reviews.

Fourth, hiding assets has become both easier and more detectable. Digital assets, foreign accounts, and tokenized holdings can be concealed, but blockchain analytics and AI-driven forensic accounting make concealment progressively riskier. Courts sanction nondisclosure severely, including awarding the concealed asset entirely to the innocent spouse in some jurisdictions. The discovery phase increasingly relies on machine-assisted review precisely because the volume of digital records in a founder's life is enormous; attempting to hide within that volume fails more often than it succeeds.

Fifth, founders sometimes transfer equity to trusts or relatives mid-divorce to shield assets. Every US jurisdiction treats such transfers during pending proceedings as voidable fraudulent conveyances, and they poison credibility on every other issue. The correct path is transparent structuring done well before marital breakdown, ideally through prenuptial or postnuptial agreements.

Timing: When to Act and What Deadlines Matter

Classification questions reward early action. Prenuptial agreements executed before marriage remain the cheapest protection available — a well-drafted agreement specifying treatment of future startup equity costs a fraction of one percent of what contested litigation over a nine-figure company costs. Postnuptial agreements work during the marriage but face heightened scrutiny for fairness and disclosure. Once separation occurs, statutory deadlines govern: many states impose rules on when valuations are 'as of' (filing date, separation date, or trial date), and choosing the valuation date strategically can shift millions when an AI company's trajectory is steep. California, for example, generally values community assets at trial unless the court directs otherwise, which benefits the spouse expecting continued appreciation; other jurisdictions freeze values earlier.

Temporary support hearings arrive within weeks of filing and set the tone. Because AI executives' income is lumpy — salary plus irregular option exercises — presenting a realistic cash-flow picture early prevents support orders built on inflated paper wealth. Discovery then runs for months; in complex cases, expert reports, depositions, and mediation typically consume 12–24 months before trial, matching the multi-year arc seen in the SK chairman proceedings. Waiting passively through this period is itself a decision: evidence goes stale, employees leave, models get superseded, and memories blur about which milestones occurred when.

For the non-founder spouse, the practical advice is to retain counsel and a forensic accountant immediately upon contemplating separation, before the founder restructures anything. For founders, the mirror-image advice applies: get a defensible independent valuation on record early, document the separate-property origins of premarital code and capital, and consider negotiating rather than litigating once expert fees alone begin approaching seven figures.

Costs, Fees, and What This Process Actually Runs

High-net-worth AI divorces are expensive at every tier. Retainers for experienced family law counsel in major markets commonly run $25,000–$100,000 upfront, with blended hourly rates of $500–$1,200 per attorney. Business valuators charge $15,000–$75,000 per engagement, and dueling experts mean both sides pay. Forensic accountants add $10,000–$50,000 depending on entity complexity. EDiscovery hosting and AI-assisted review for a data-heavy case can run $5,000–$50,000 or more, though machine-assisted review typically reduces total review spend by 30–60% compared to linear manual review of equivalent volumes. Total contested litigation for a case involving a nine-figure private company routinely exceeds $500,000 per side and can surpass $2 million when trials run multiple weeks.

Against that backdrop, mediation and collaborative divorce offer meaningful savings — often resolving in 6–12 months at 20–40% of litigated cost — provided both parties disclose fully and the valuation gap is bridgeable. AI-enabled drafting and research tools used by counsel reduce document-drafting hours, and some of those savings reach clients, though the technology assists rather than replaces judgment on classification and structure decisions. The cheapest outcome, by orders of magnitude, remains a prenuptial agreement drafted before any wealth exists: $5,000–$25,000 in fees versus seven figures of later conflict.

One caution cuts against uncritical enthusiasm for automation: valuation ultimately requires judgment about markets, management quality, and technology durability that no software reliably supplies. Tools excel at document processing, timeline construction, and anomaly detection; they do not replace a credentialed appraiser testifying under cross-examination. Parties who treat AI outputs as authoritative valuations rather than inputs invite errors that opposing experts will exploit.

The Bottom Line for Anyone Facing an AI-Equity Divorce

Marital versus separate property analysis of AI-built wealth rests on familiar doctrines — acquisition timing, commingling, active versus passive appreciation, vesting apportionment — applied to assets whose value is unusually volatile, illiquid, and evidence-intensive. The SK chairman's billion-dollar ruling, the Two Sigma trial coverage, and the growing literature on founders' divorces all point to the same operational conclusions: map every equity instrument and deferred-compensation promise, date the value creation rigorously using technical and financial records, choose valuation methods and settlement structures deliberately rather than defaulting to whatever the first expert proposes, and settle structurally when possible because litigation outcomes in this domain are genuinely uncertain. Both founders and non-founder spouses benefit from moving early, disclosing completely, and spending on quality valuation expertise rather than scorched-earth motion practice. The technology changes; the fiduciary discipline of dividing it does not.