The Suicide Region: Option Games and the Race to Artificial General Intelligence

Abstract: Standard real options theory predicts delay in exercising the option to invest or deploy when extreme asset volatility or technological uncertainty are present. However, in the current race to develop artificial general intelligence (AGI), sovereign actors are exhibiting behaviors contrary to theoretical predictions: the US and China are accelerating AI investment despite acknowledging the potential for catastrophic failure from AGI misalignment. We resolve this puzzle by formalizing the AGI race as a continuous-time preemption game with endogenous existential risk. In our model, the cost of failure is no longer bounded only by the sunk cost of investment (I), but rather a systemic ruin parameter (D) that is correlated with development velocity and shared globally. As the disutility of catastrophe is embedded in both players' payoffs, the risk term mathematically cancels out of the equilibrium indifference condition. This creates a "suicide region" in the investment space where competitive pressures force rational agents to deploy AGI systems early, despite a negative risk-adjusted net present value. Furthermore, we show that "warning shots" (sub-existential disasters) will fail to deter AGI acceleration, as the winner-takes-all nature of the race remains intact. The race can only be halted if the cost of ruin is internalized, making safety research a prerequisite for economic viability. We derive the critical private liability threshold required to restore the option value of waiting and propose mechanism design interventions that can better ensure safe AGI research and socially responsible deployment.

• The paper demonstrates that rational actors preemptively deploy AGI even with negative NPV due to the shared cost of systemic ruin.
• It models AGI development as a continuous-time option game, revealing the failure of classical deterrence in the face of unbounded catastrophic risk.
• It proposes policy interventions—such as internalizing externalities and windfall clauses—to shift deployment thresholds and ensure safer AGI progress.

The Suicide Region in AGI: Option Games, Preemption, and Existential Risk

Introduction

This paper rigorously formalizes the strategic dynamics underlying the geopolitical race towards AGI, highlighting the paradoxical acceleration observed among sovereign actors despite acknowledged existential risks. Contradicting canonical real options theory, which would typically predict delay in irreversible investments under uncertainty, the model demonstrates that rational actors are incentivized to preemptively deploy AGI—even with negative risk-adjusted net present value (NPV)—because the cost of systemic ruin is a shared and unavoidable externality. This equilibrium phenomenon is characterized as the "suicide region." The analysis draws on and extends option games in continuous time with endogenous existential risk, presenting critical theoretical implications for both economic modeling and AGI safety governance.

Theoretical Model and Core Insights

Standard real options theory (Dixit & Pindyck, 1994; Weeds, 2002) prescribes deferral of irreversible investments when volatility and uncertainty are extreme, creating substantial value in waiting for information resolution. In R&D races, competitive dynamics can lead to preemptive investment but only up to the point where increased volatility induces a "sleeping patent" equilibrium. However, the model constructed here incorporates the AGI context, where first-mover advantage is absolute (winner-takes-all), and downside risk is unbounded, incorporating a monetized systemic ruin parameter $D$ that reflects existential catastrophe.

The game is defined as a two-agent, continuous-time stochastic process over the asset value of AGI development, featuring a probability $\pi(\tau)$ of alignment that is an increasing, concave function of safety research time $\tau$. Payoff structures account for the likelihood of success (alignment), economic rents (asymptotically zero for the follower), and systemic catastrophe.

A pivotal result is the "cancellation effect" in the indifference condition: Since the cost of systemic ruin appears identically in both the leader and follower payoffs, it algebraically cancels out. Thus, the risk of global catastrophe is not incorporated in the equilibrium deployment threshold—it fails to act as a deterrent. When competitive preemption incentives outweigh the rational NPV condition ($L > F$ but $L < 0$), actors accelerate AGI deployment into the "suicide region," characterized by strategies that rationalize deployment even with expected negative payoff.

Failure of Deterrence: Contrasts with Classical Security Studies

The analogy with nuclear deterrence—where mutually assured destruction (MAD) raises the threshold for first strike in direct proportion to the magnitude of D—breaks down in the AGI context. There is no credible second-strike mechanism in AGI; winner-take-all dynamics yield S (the follower’s market share) asymptotically approaching zero. Therefore, the magnitude of systemic ruin does not deter early exercise; the only rational hedge is to be the first mover, since being the second mover does not mitigate exposure to existential risk.

Benevolent Preemption and the Saviour's Trap

The model incorporates asymmetric beliefs in safety alignment, structurally capturing the empirically observed "saviour’s trap" or "unilateralist’s curse." Actors believing themselves strictly safer than their rivals are even more incentivized to accelerate deployment to preclude a presumably less safe rival from deploying first. This benevolent motivation creates a "saviour premium" in the preemption calculus: As perceived misalignment risk rises for the rival, the threshold for preemptive deployment further decreases, paradoxically raising systemic risk.

Policy Mechanisms and Equilibrium Restructuring

The "suicide region" persists unless the underlying payoff structures are modified. The paper delineates several intervention strategies:

• Internalization of Externality: Sufficiently privatizing systemic ruin through direct liability (e.g., catastrophe bonds, escrow, or strict tort liability) can shift the preemption threshold to restore the value of waiting. When the leader is exposed to a critical level of private risk, the rational strategy reverts to waiting until NPV is positive, halting the AGI race in the suicide region.
• Windfall Clauses / Market Share Redistribution: If post-AGI rents are even partially socialized (i.e., S is significantly above zero), the incentive for reckless preemption decays. Windfall treaties that allocate benefits to both first and second movers defang the "winner-takes-all" property, increasing rational deployment thresholds and incentivizing heightened safety.
• Strategic Uncertainty and Monitoring: The inability to verify the rival’s development status incentivizes racing; robust, verifiable monitoring mechanisms could suppress early preemption during periods when neither actor is near deployment capability. However, mirroring the "breakout" phenomenon in arms control, perfect information may induce instability in late-stage development.

Importantly, "warning shots" (localized or sub-existential catastrophes) do not alter the equilibrium unless they instantiate compensatory changes in private liability or rent sharing. Contrary to common expectations, exogenous shocks will not, by themselves, resolve the coordination puzzle or restore the value of delay.

Implications for AGI Safety and International Governance

The analysis demonstrates that voluntary moratoriums or pause agreements are structurally insufficient. Only by fundamentally altering liability structures or benefits allocation can the pathological Nash equilibrium be averted. These findings are salient for AGI governance, as the model predicts that, absent global enforcement of risk internalization or benefit distribution, rational actors will circumvent safety protocols regardless of catastrophic tail risks. Mechanism design that mandates safety research, liability posting, or windfall redistribution is necessary to force alignment between rational economic decisions and collective human survival.

Conclusion

This work decisively extends the economic theory of real options in the explicit context of endogenous existential risk, making three principal contributions:

1. It demonstrates, through rigorous equilibrium analysis, why existential risk fails as a deterrence mechanism in the AGI arms race under preemption,
2. It delineates the parameter space ("suicide region") in which rational actors will deploy AGI prematurely, and
3. It specifies mechanism design interventions that could restore rational delay and enable safe AGI deployment.

Absent institutional reforms that privatize catastrophic risk or socialize AGI rents, the current competitive equilibrium inexorably leads to premature, unsafe deployment—a Nash equilibrium from which voluntary restraint or warning events will not extricate the system. The theoretical and policy implications of this model have direct bearing on the viability of AGI safety governance and the broader ethics of technological acceleration.