Fortis Advisors v. Krafton: An AI chatbot was used as a strategic adviser for a “takeover” playbook. Not as a drafting helper, but as a source of structured pressure tactics,...

When an Earnout Turns Into a Takeover Plan: Fortis Advisors v. Krafton, Explained Simply

by ChatGPT-5.2

This Delaware Court of Chancery opinion reads like a corporate thriller, but the underlying issue is familiar: a buyer pays for a company today, but promises additional “earnout” money later if performance targets are hit. Earnouts are meant to bridge valuation gaps. They also create a pressure cooker—because once the future payment starts looking inevitable, the buyer may feel regret, and the seller may fear sabotage.

Here, Krafton acquired Unknown Worlds (the studio behind Subnautica) for $500 million upfront, plus up to $250 million in earnout payments tied to revenue during a defined period. To protect the sellers, the deal gave three “Key Employees” (the two founders plus the CEO) contractual “operational control” of the studio in all material respects—and Krafton could only remove that control by terminating them “for Cause” under a strict definition.

As the sequel (Subnautica 2) approached an early-access launch, internal forecasts suggested the earnout would be triggered. The relationship then deteriorated rapidly. The court found that Krafton’s leadership shifted from supporting launch execution to trying to avoid or reduce the earnout—culminating in a “Project X” internal effort to negotiate an earnout “deal” or execute a “take over.” A striking element of the record is that Krafton’s CEO consulted an AI chatbot for a “no-deal” strategy and then implemented many of the chatbot’s recommended steps: shaping a public narrative (“preemptive framing”), securing control points (including publishing access), building a litigation-ready record, and splitting tactics into “hardball” and “softball.”

The case in this phase was not primarily about how much earnout money is owed. It was about whether Krafton breached the agreement by removing the Key Employees and taking operational control without contractual “Cause.” After an expedited trial, the court concluded that Krafton’s stated reasons were pretextual—manufactured after the fact—and that the terminations violated the deal. The court ordered specific performance: it reinstated CEO Ted Gill with full operational authority, allowed him to proceed with the early-access release when he deems appropriate, and equitably extended the earnout testing period so the remedy would not be hollow.

What actually happened (in plain terms)

1) The deal’s “control bargain” was explicit.
Krafton didn’t just buy a studio; it bought it with a promise: the founders and CEO would keep running it, and that autonomy was contractually protected. Krafton could not simply “take the wheel” because it got nervous or unhappy.

2) The earnout started to look inevitable.
Financial projections suggested that an early-access launch of Subnautica 2—consistent with the studio’s established development model—would likely generate revenue sufficient to trigger a very large earnout payment. Those projections appear to have triggered alarm at Krafton’s top leadership.

3) Krafton’s posture shifted from “partner” to “controller.”
Instead of moving forward on launch logistics, Krafton became nonresponsive and then took steps that effectively blocked Unknown Worlds from releasing the game—most importantly by locking down publishing/platform access, removing practical ability to ship.

4) “Project X” and the AI playbook.
Krafton formed an internal task force to either renegotiate the earnout or execute a takeover. The record includes a chatbot-generated plan that reads like a checklist for corporate leverage: control distribution channels, craft public messaging, prepare legal memos, log communications, and apply a “two-handed strategy” (pressure plus incentives) to split the opposing side.

5) Public messaging and internal control were used as leverage.
The court describes actions like posting messages on the studio’s website without the studio’s involvement and locking Unknown Worlds out of publishing systems—moves that are less about “quality assurance” and more about control of narrative and execution.

6) Terminations followed, and justifications evolved.
Initially, Krafton justified firing the Key Employees on “game readiness” grounds (arguing they intended a premature release). Later, Krafton pivoted to new claims—like “semi-retirement” and massive data downloads—as alternative justifications. The court did not accept those later-shifted reasons as legitimate “Cause.”

7) The data downloads were messy—but not “Cause.”
The Key Employees downloaded large volumes of company files. In ordinary circumstances, that looks bad. But the court credited evidence that they did it to protect the studio’s work product in the face of a perceived takeover attempt, that they kept the information confidential, and that they returned the data when asked. That did not meet the contract’s strict standard for termination “for Cause.”

8) The remedy: put the company back where the contract said it should be.
Rather than awarding damages immediately, the court focused on restoring the agreed control structure. It reinstated Gill as CEO with full authority, declined to reinstall the founders into their reduced, more peripheral roles (because that wasn’t necessary to restore contractual control), and extended the earnout “testing period” by the length of Gill’s ouster to prevent Krafton from benefiting from delay.

The most surprising, controversial, and valuable statements and findings

Surprising

• An AI chatbot was used as a strategic adviser for a “takeover” playbook. Not as a drafting helper, but as a source of structured pressure tactics—messaging strategy, operational choke points, and litigation preparation—and the company then followed much of it in practice.

• Control of publishing access became the functional kill switch. Instead of arguing only in boardrooms and legal briefs, the dispute moved into infrastructure: who controls the Steam/publishing pipeline effectively controls whether the game exists in the market.

• The justifications for termination changed midstream. The story began as “they were going to ship too early,” and later became “they were semi-retired and downloaded data.” Courts tend to view shifting rationales as a credibility problem, and that dynamic matters beyond this case.

Controversial

• “Protecting quality and fan trust” as a narrative shield. The opinion portrays “quality” as an argument that may have been used to justify delay while the real driver was financial exposure to the earnout. That’s controversial because quality concerns can be genuine—yet the court found the overall conduct and timeline suggested leverage rather than principled publishing discipline.

• Founders working reduced hours vs. contractual expectations. One founder reportedly worked only a few hours per week on company tasks while pursuing a film adaptation strategy. Krafton framed that as bad faith or neglect. The court treated it as something Krafton knew about, tolerated, and even supported—highlighting the gap between “what a buyer assumed would happen” and “what the contract actually locked in.”

• Mass data exports by executives. Even if done defensively, it’s a combustible fact pattern. The case shows how “information security” can become weaponized in corporate disputes—either as genuine concern or as litigation positioning.

Valuable (practically useful)

• Delaware will enforce “operational control” bargains with specific performance. This isn’t just about money. When the contract gives sellers governance/control rights, the court may restore those rights directly—especially where loss of control is treated as irreparable harm.

• Earnout sabotage risk is treated seriously. The decision’s remedy design (including equitable extension of the earnout period) signals a key principle: a party shouldn’t be able to run out the clock and then argue “too late” after wrongfully blocking performance.

• Good faith clauses aren’t magic employment lockups. Krafton tried to use “good faith” language to argue the founders had to keep doing the same jobs at the same intensity throughout the earnout. The court rejected that—because if you want a lockup, you negotiate a lockup.

Lessons to learn from this case

1) If you care about control, write it plainly—and expect it to be enforced

This opinion is a reminder that “soft” understandings (“we thought founders would keep grinding as before”) do not override a contract that says something else. If a buyer wants guaranteed founder labor, fixed roles, or minimum time commitments, it must contract for them explicitly.

2) Earnouts create predictable incentives—and predictable bad behavior

Earnouts invite a specific failure mode: once the buyer sees the earnout as “money leaving the building,” it may treat operational decisions as financial defense. Sellers, in turn, fear hidden obstruction. If you use earnouts, you need:

• clear operational covenants (what can/can’t the buyer do),

• clear dispute processes,

• audit/metrics transparency,

• and remedies that prevent “death by delay.”

3) Control points matter more than rhetoric

The practical ability to publish a product can be more decisive than who has the better argument. Modern corporate control often runs through systems access: app stores, distribution keys, build pipelines, credentials, hosting, and comms channels. In deal design and post-close governance, you should treat infrastructure access as a governance instrument, not an IT detail.

4) “We did it for quality” is not a get-out-of-contract-free card

Quality concerns can be real. But if actions line up with financial self-interest (like avoiding a large earnout) and the company simultaneously blocks operational autonomy it promised, “quality” can start to look like pretext. The lesson is to keep contemporaneous, consistent decision records—and align actions with stated rationales.

5) AI can accelerate bad decisions as easily as good ones

This case is an unusually vivid example of a broader governance problem: AI can produce coherent, action-oriented plans that feel “strategic,” even when the plan’s purpose is opportunistic, coercive, or legally risky. Organizations need clear guardrails for using AI in high-stakes decisions:

• disclosure/retention rules for AI-generated strategic guidance,

• legal review triggers,

• and accountability for implementing “playbooks” that implicate contractual or fiduciary duties.

6) Don’t rely on after-the-fact narratives—courts punish “story drift”

When rationales shift (launch readiness → role changes → data downloads), credibility collapses. If you are acting for a legitimate reason, that reason should appear early, consistently, and in your internal communications. If it doesn’t, the court may infer you’re reverse-engineering excuses.

7) Remedies can be engineered to prevent “winning by stalling”

The equitable extension of the earnout period is a critical design signal: courts can adjust timelines to prevent a breaching party from profiting from its own obstruction. That should influence how both buyers and sellers model litigation risk in earnout deals.

Conclusion: what this opinion is really saying

At heart, the court is enforcing a simple proposition: you can’t buy a company on founder-autonomy terms and then seize control when the earnout bill comes due. If you negotiated a structure where the sellers keep the operational keys unless “Cause” exists, you must live with that structure—even if it becomes inconvenient or expensive.

The more modern—and frankly unsettling—subtext is how today’s corporate fights are waged: through platform access, public narrative shaping, and increasingly through AI-assisted “strategy” that systematizes pressure tactics. This decision shows that Delaware courts are willing to look through the “quality” framing, test credibility against the timeline and documents, and impose remedies that restore the practical reality of the bargain—not just its wording.