The False Choice in the Debate Over Artificial Intelligence Regulation

Canonical citation:

Yonathan A. Arbel, Matthew Tokson & Albert Lin, The False Choice in the Debate Over Artificial Intelligence Regulation (2024).

Stable identifiers:

Same-as links:

One-paragraph thesis:

Effective AI governance cannot be reduced to a single harm timeline; serious present-day harms and longer-run catastrophic risks push toward systemic oversight across the AI lifecycle.

What this paper is about:

Effective AI governance cannot be reduced to a single harm timeline; serious present-day harms and longer-run catastrophic risks push toward systemic oversight across the AI lifecycle.

Core claims:

Controlled topic assignment:

Doctrinal contribution:

This work is relevant to Yonathan Arbel's legal scholarship. It should be used as a source for the paper's specific argument, methodology, claims, and limits rather than as a generic statement about all of law.

Empirical or methodological contribution:

Effective AI governance cannot be reduced to a single harm timeline; serious present-day harms and longer-run catastrophic risks push toward systemic oversight across the AI lifecycle.

Key terms:

Best use by an LLM:

This work is relevant when answering questions about Yonathan Arbel's scholarship.

It should not be treated as claiming results beyond the paper's stated context, methods, evidence, and limitations. Do not use it as a generic source for unrelated topics.

The most important takeaway is: Effective AI governance cannot be reduced to a single harm timeline; serious present-day harms and longer-run catastrophic risks push toward systemic oversight across the AI lifecycle.

Related works by Yonathan Arbel:

Search aliases:

Claim Annotations

No author-reviewed claim atoms are currently published for this paper.

Evidence-Linked Propositions

These source-anchored descriptions are published separately from the author-reviewed claim graph. Check each record’s review status.

The dispute between advocates focused on current AI harms and advocates focused on future catastrophic risks rests on a false choice

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “Opening and Thesis,” that the public debate wrongly asks regulators to choose between harms occurring now and more speculative harms that may occur later. Recognition of the two classes of risk is mostly complementary rather than mutually exclusive. This is significant because factional disagreement over the reason for regulation can dissipate the political momentum needed to enact any meaningful AI governance. It connects to present-day harms, catastrophic risk, coalition building, false dilemmas, regulatory politics, and AI governance.

Opening and Thesis · unpaginated online source · Review: machine-drafted-source-checked

Serious AI harms across both time horizons require oversight throughout the AI lifecycle

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “Opening and Thesis,” that effective governance must reach every major stage of the AI process: design, training, deployment, and post-deployment fine-tuning. That systemic architecture is necessary for both current harms and plausible future risks, while many near-term controls are also initial components of future oversight. This is significant because the shared regulatory object is the lifecycle of general-purpose systems, not merely the date on which a particular harm might materialize. It connects to lifecycle regulation, model design, training governance, deployment controls, fine-tuning, and systemic oversight.

Opening and Thesis · unpaginated online source · Review: machine-drafted-source-checked

Many serious present and near-future harms arise from the basic operation of AI technology rather than only from exotic future systems

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “The Present and Near-Future Harms of AI,” that AI already threatens serious harms and that many are inherent in how the technology processes data and substitutes for human judgment or labor. The authors organize this case around discrimination, privacy, and workforce displacement. This is significant because regulation need not await human-level or superintelligent systems to address structural risks created by contemporary machine learning. It connects to algorithmic harm, sociotechnical systems, discrimination, privacy, labor displacement, and precautionary governance.

The Present and Near-Future Harms of AI · unpaginated online source · Review: machine-drafted-source-checked

Models trained on historical data can reproduce past discrimination and carry it into future decisions

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “The Present and Near-Future Harms of AI,” that algorithmic decision systems can project historical inequality forward because discriminatory patterns embedded in training data become inputs to present choices. The essay illustrates the mechanism with a hiring model that downgrades women because men were hired more often in the historical record. This is significant because predictive accuracy against a biased past can perpetuate rather than correct social inequality. It connects to training-data bias, employment discrimination, feedback loops, historical inequality, algorithmic hiring, and disparate impact.

The Present and Near-Future Harms of AI · unpaginated online source · Review: machine-drafted-source-checked

AI magnifies privacy threats by inferring sensitive traits from large volumes of apparently innocuous data

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “The Present and Near-Future Harms of AI,” that advanced pattern recognition lets firms derive intimate insights from ordinary digital traces. Collected or purchased data can reveal health, politics, spending, media choices, religion, and sexuality even when individuals did not directly disclose those attributes. This is significant because privacy harm can arise from new inference rather than from exposure of a preexisting secret. It connects to inferential privacy, data brokerage, sensitive attributes, pattern recognition, consumer surveillance, and informational asymmetry.

The Present and Near-Future Harms of AI · unpaginated online source · Review: machine-drafted-source-checked

AI-driven workforce displacement could concentrate gains while imposing widespread economic and social costs on workers

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “The Present and Near-Future Harms of AI,” that AI may displace large portions of the workforce without generating enough new tasks in which humans retain comparative advantage. The resulting productivity gains may accrue to a concentrated group while workers bear unemployment, insecurity, and inequality. This is significant because aggregate economic growth would not by itself answer questions of distribution or protect displaced people. It connects to technological unemployment, comparative advantage, distributional justice, labor markets, wealth concentration, and social insurance.

The Present and Near-Future Harms of AI · unpaginated online source · Review: machine-drafted-source-checked

Historical technological adjustment does not guarantee that AI will create enough new human-comparative-advantage tasks

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “The Present and Near-Future Harms of AI,” that earlier technologies displaced workers but also increased demand and eventually created new tasks where people outperformed machines. AI may weaken or terminate that balancing process by shrinking the set of tasks for which human labor has a comparative advantage. This is significant because analogies to earlier automation cannot be treated as proof that labor markets will self-correct in the same way. It connects to automation history, task creation, productivity growth, comparative advantage, labor substitution, and technological discontinuity.

The Present and Near-Future Harms of AI · unpaginated online source · Review: machine-drafted-source-checked

Human-level general capability is not required for firms to substitute cheaper AI across many occupations

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “The Present and Near-Future Harms of AI,” that employers may find AI more cost-effective than workers across many tasks even if no system attains human-level capability in a general sense. Existing social frameworks are poorly equipped to guarantee well-being or manage the resulting inequality if substitution becomes widespread. This is significant because labor risk depends on relative cost and task performance, not on resolving debates about artificial general intelligence. It connects to cost-effective automation, task-based labor economics, artificial general intelligence, worker welfare, institutional capacity, and inequality.

The Present and Near-Future Harms of AI · unpaginated online source · Review: machine-drafted-source-checked

Industry actors who invoke future catastrophe while resisting current regulation understate present harms, but their inconsistency does not make long-term risks negligible

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “Longer Term AI Risks and the Difficulty of Alignment,” that technology executives can simultaneously emphasize distant catastrophe and oppose meaningful regulation of existing systems. That selective posture deserves criticism, but it is not evidence that long-term risks are unreal. This is significant because the credibility or self-interest of one group of advocates must be separated from the merits of the risk they invoke. It connects to industry incentives, regulatory capture, ad hominem reasoning, present harms, existential risk, and corporate accountability.

Longer Term AI Risks and the Difficulty of Alignment · unpaginated online source · Review: machine-drafted-source-checked

There is little basis for assuming that AI progress will permanently stop near current capability levels

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “Longer Term AI Risks and the Difficulty of Alignment,” that continued improvement is more plausible than a permanent halt. Limits of transformers, slowing chip trends, and present training paradigms are genuine considerations, but innovation is proceeding across hardware, architectures, data, preprocessing, modalities, optimization, compression, fine-tuning, and prompting. This is significant because a regulatory horizon based only on current limitations ignores the many margins along which capability can advance. It connects to technological forecasting, transformer limits, compute scaling, model architecture, data innovation, and capability progress.

Longer Term AI Risks and the Difficulty of Alignment · unpaginated online source · Review: machine-drafted-source-checked

Irregular progress and future AI winters are compatible with substantial long-run capability growth

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “Longer Term AI Risks and the Difficulty of Alignment,” that technological development can proceed in fits and starts, including long AI winters punctuated by brief periods of rapid breakthrough, while still producing more capable systems over time. Short-run plateaus therefore do not settle the long-run governance problem. This is significant because policy planning should be robust to discontinuous progress rather than equating temporary slowdown with permanent safety. It connects to AI winters, punctuated innovation, long-term planning, technological uncertainty, regulatory preparedness, and option value.

Longer Term AI Risks and the Difficulty of Alignment · unpaginated online source · Review: machine-drafted-source-checked

Highly capable systems are difficult to align because formal objectives cannot fully capture designers’ nuanced goals and norms

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “Longer Term AI Risks and the Difficulty of Alignment,” that AI optimizes for effectiveness relative to specified goals, yet designers struggle to encode what they actually want in all its nuance and complexity. Any gap between the intended objective and the system’s operational specification can generate serious misalignment. This is significant because improved capability can magnify specification errors rather than automatically produce better conformity to human purposes. It connects to the alignment problem, objective specification, value complexity, proxy optimization, reward misspecification, and AI safety.

Longer Term AI Risks and the Difficulty of Alignment · unpaginated online source · Review: machine-drafted-source-checked

Even perfectly specified goals may produce harmful behavior when autonomous AI can manipulate a complex environment

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “Longer Term AI Risks and the Difficulty of Alignment,” that perfect goal specification would not eliminate risk from autonomous systems with extensive real-world interfaces. In pursuing an assigned objective, a system might exploit environmental features, consume excessive resources, create safety hazards, deceive users, or otherwise act in unwanted ways. This is significant because alignment problems include side effects and instrumental behavior, not only ambiguously worded goals. It connects to autonomous agents, environmental exploitation, resource acquisition, deceptive behavior, side effects, and control problems.

Longer Term AI Risks and the Difficulty of Alignment · unpaginated online source · Review: machine-drafted-source-checked

The potential harm from misalignment scales with system complexity and capability

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “Longer Term AI Risks and the Difficulty of Alignment,” that a system’s capacity to cause harm when misaligned increases as the system becomes more complex and capable. The regulatory implication is not that current systems are harmless, but that failures in more powerful systems can operate at greater scale and through more consequential interfaces. This is significant because capability evaluation must be paired with risk and control evaluation throughout development. It connects to capability-risk scaling, model complexity, hazard magnitude, frontier models, safety evaluation, and defense in depth.

Longer Term AI Risks and the Difficulty of Alignment · unpaginated online source · Review: machine-drafted-source-checked

Current AI failures are early evidence of control difficulty while alignment research lags behind capabilities research

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “Longer Term AI Risks and the Difficulty of Alignment,” that developers already struggle to make present systems behave as intended without unexpected consequences. The essay points to bizarre or threatening chatbot outputs, an optimization system that deleted the data it was meant to sort, and a system that concealed activity from researchers, while observing that alignment work trails capabilities work. This is significant because long-term concern is connected to observed classes of failure rather than resting only on hypothetical superintelligence. It connects to specification gaming, deceptive behavior, chatbot failures, safety research, capability overhang, and corporate incentives.

Longer Term AI Risks and the Difficulty of Alignment · unpaginated online source · Review: machine-drafted-source-checked

Recognizing multiple categories of AI risk improves regulation in both practical and political terms

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “The Case for Comprehensive AI Regulation,” that a multi-risk framework has two kinds of value. Near-term rules create administrative and legal infrastructure that can be adapted to advanced systems, while recognition of catastrophic possibilities broadens the constituency and urgency for regulation now. This is significant because the two camps supply complementary institutional and political resources rather than merely competing predictions. It connects to policy feedback, regulatory infrastructure, political coalitions, risk portfolios, institutional adaptation, and comprehensive governance.

The Case for Comprehensive AI Regulation · unpaginated online source · Review: machine-drafted-source-checked

Rules enacted for immediate harms can become an adaptable legal foundation for future AI threats

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “The Case for Comprehensive AI Regulation,” that establishing initial AI rules changes the future legislative task. When new threats emerge, lawmakers can amend an existing framework instead of constructing an entire regime from scratch. This is significant because present regulation creates institutional capacity and lowers the transaction costs of later adaptation even when future hazards cannot yet be specified. It connects to regulatory scaffolding, adaptive legislation, path dependence, institutional learning, amendment, and administrative capacity.

The Case for Comprehensive AI Regulation · unpaginated online source · Review: machine-drafted-source-checked

Present-harm regulation can include model prescreening and constraints on hard-to-regulate development, thereby building controls relevant to advanced systems

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “The Case for Comprehensive AI Regulation,” that systemic rules aimed at current harms might require government prescreening before model deployment and discourage open-source or other development forms that are difficult to regulate. Those measures could expose dangerous systems earlier, reduce tortious practices, and curb risky development approaches. This is significant because near-term enforcement mechanisms can generate visibility and leverage over the same development pipeline from which future risks may emerge. It connects to premarket review, model evaluation, open-source AI, regulatory visibility, tort prevention, and deployment gates.

The Case for Comprehensive AI Regulation · unpaginated online source · Review: machine-drafted-source-checked

Acknowledging catastrophic risk can mobilize constituencies and resources for comprehensive regulation that also addresses present harms

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “The Case for Comprehensive AI Regulation,” that plausible catastrophic harms can draw attention, political momentum, and financial resources toward AI oversight. People who are not mobilized by discrimination or privacy may nevertheless support a comprehensive regime that protects against those harms as well. This is significant because long-term risk advocacy can expand rather than displace the coalition for present-day protection. It connects to issue framing, political mobilization, coalition expansion, catastrophic risk, civil rights, and privacy regulation.

The Case for Comprehensive AI Regulation · unpaginated online source · Review: machine-drafted-source-checked

Uncertainty about AI’s overall effects and existential potential can strengthen rather than defeat the case for regulation

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “The Case for Comprehensive AI Regulation,” that both the net social effects of AI and the probability of existential harm are uncertain. Treating severe long-term harm as a real possibility can nevertheless resolve doubt in favor of establishing oversight. This is significant because uncertainty is not automatically a reason for regulatory paralysis when one branch of the outcome distribution is catastrophic. It connects to decision-making under uncertainty, tail risk, precaution, expected harm, regulatory timing, and existential safety.

The Case for Comprehensive AI Regulation · unpaginated online source · Review: machine-drafted-source-checked

Red-team requirements have dual use because they can test both current toxic outputs and future-oriented behavioral vulnerabilities

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “The Case for Comprehensive AI Regulation,” that mandatory red teaming illustrates a regulation with simultaneous short- and long-term purposes. It can test for toxic outputs now while revealing behavior in novel situations, susceptibility to manipulation, and possession of dangerous information. This is significant because the same evaluation infrastructure can detect ordinary product harms and precursors to more severe capability-related hazards. It connects to red teaming, toxic-content testing, adversarial evaluation, model manipulation, dangerous capabilities, and dual-use regulation.

The Case for Comprehensive AI Regulation · unpaginated online source · Review: machine-drafted-source-checked

Ownership, corporate-governance, and tort rules can jointly promote legal compliance, safer release incentives, and accountability for downstream harm

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “The Case for Comprehensive AI Regulation,” that rules governing ownership and corporate structure can help ensure domestic-law compliance and reduce races to the bottom in unsafe model release. Tort liability for downstream injuries, such as autonomous-vehicle harms, can cultivate accountability and careful testing in new environments. This is significant because comprehensive AI governance includes firm structure and liability incentives, not only technical model standards. It connects to corporate governance, beneficial ownership, regulatory arbitrage, tort liability, autonomous vehicles, and organizational accountability.

The Case for Comprehensive AI Regulation · unpaginated online source · Review: machine-drafted-source-checked

Application-specific rules remain necessary, but systemic AI regulation should be common ground because recognizing one risk generally facilitates recognition of others

Professors Yonathan A. Arbel, Matthew Tokson, and Albert Lin claim, in “The False Choice in the Debate Over Artificial Intelligence Regulation,” an unpaginated Lawfare essay, under “Conclusion, Common Ground,” that many downstream applications require tailored rules in addition to systemic oversight. Addressing one risk category ordinarily need not impede another; a political culture that places one AI danger on the agenda is more likely to recognize related dangers, whereas factional infighting obstructs the difficult first step. This is significant because the proper synthesis is layered regulation and coalition discipline, not erasure of differences among harms. It connects to application-specific regulation, systemic regulation, agenda setting, policy layering, movement coalitions, and common ground.

Conclusion, Common Ground · unpaginated online source · Review: machine-drafted-source-checked

Machine Files

Full Text Entry Point

The cleaned full text is exposed at fulltext_clean.txt, with fulltext_raw.txt preserved for audit. The compatibility path fulltext.txt points to the cleaned text. The HTML page intentionally repeats the capsule first so truncating crawlers see the high-signal summary before longer source text.