Key Takeaways
- The 2024 Loper Bright Enterprises v. Raimondo decision overturned the Chevron doctrine, ending automatic judicial deference to agency interpretations of ambiguous statutes.
- Although Loper Bright appears to curtail agency power, scholars argue it is substantively similar to Chevron and may still allow agencies to exert preemptive influence over state tort law, especially in emerging fields like artificial intelligence (AI).
- Preemption in AI governance risks suppressing the information‑producing function of state‑level litigation, which serves as a laboratory for policy experimentation.
- A persuasive‑reasoning approach—modeled after Skidmore v. Swift & Co.—would require agencies to justify their interpretations, inviting scrutiny and feedback from state tort suits while preserving a uniform regulatory floor.
- Relying on AI or algorithmic tools for agency decision‑making creates a new form of deference (“algorithmic deference”) that lacks transparency and accountability; encoded policy should receive no judicial deference.
- Ultimately, democratic accountability in AI regulation depends on agencies that can convince through reasoned explanation rather than merely commanding through authority.
The Shift from Chevron Deference to Post‑Loper Bright Scrutiny
Since the U.S. Supreme Court overruled Chevron v. Natural Resources Defense Council in Loper Bright Enterprises v. Raimondo in 2024, legal scholars have debated the implications for the separation of federal powers. The Court’s holding “stopped federal agencies from interpreting for themselves the limits of their regulatory authority, thereby preventing them from declaring the rules to be preemptive regulatory ceilings.” In other words, judges are no longer compelled to accept an agency’s reading of an ambiguous statute simply because it is reasonable; they must now evaluate whether the agency’s interpretation falls within the authority delegated by Congress.
Why Chevron’s Deference Was Problematic for AI Governance
Under Chevron, courts were often compelled to accept an agency’s interpretation of an ambiguous statute so long as that interpretation was reasonable. This “judicial acquiescence” allowed federal agencies to set AI standards and simultaneously declare those standards as preemptive ceilings that barred state tort liability. Critics argued that such deference deprived states of their role as “laboratories of democracy,” a concept Justice Louis Brandeis famously described as enabling policy experimentation. By cutting off the ground‑truth signals generated by state‑level litigation, Chevron‑style deference risked producing rigid, ex ante regulations that lacked persuasive, technically sound justifications—particularly troubling in AI, where risks are probabilistic and emergent.
Loper Bright: A Superficial Break with Continuing Concerns
Although Loper Bright appears to mark a departure from Chevron, commentators such as Cary Coglianese and David B. Froomkin label the ruling “disingenuous.” They note that long before Loper Bright, decisions like United States v. Mead had already limited Chevron deference to statutes showing a clear congressional intent to delegate authority. Moreover, the Court’s interpretation of the Administrative Procedure Act (APA) treats the APA’s grant of interpretive authority to agencies as an invitation for courts to substitute their own policy preferences for agency judgments. This approach may not yield better‑informed regulations, especially in novel, technical domains like AI, where agency expertise remains essential.
The Information Suppression Versus Information Production Dilemma
The core tension in AI preemption is not merely about federal supremacy but about whether the system suppresses or produces information. State tort suits generate dynamic feedback that helps regulators calibrate policy amid shifting technological realities. When federal agencies can preempt those suits without robust justification, the feedback loop is broken, increasing the risk of ill‑fit regulations. As the article observes, “preemption is not just about federal supremacy, but about information suppression versus information production.” A regulatory framework that allows state litigation to proceed preserves the democratic benefits of experimentation while still providing a baseline of federal standards.
A Persuasive‑Reasoning Alternative: The Agency Reference Model
Catherine Sharkey’s “agency reference” model offers a promising path forward. Under this approach, federal regulations grounded in persuasive reasoning would establish a uniform regulatory floor, leaving room for state tort law to impose additional liability. By treating preemption as a floor rather than a ceiling, state tort suits can continue to operate, using judicial discovery to extract useful information that feeds back into federal rulemaking. This model aligns with the principle that agencies must articulate their reasoning, enabling others to contest interpretations and iteratively improve policy—especially vital given the information asymmetry between federal regulators and AI developers.
Returning to Skidmore‑Style Deference
In the post‑Loper Bright landscape, courts should treat agency interpretation as persuasive rather than binding, echoing the deference standard set forth in Skidmore v. Swift & Co. Under Skidmore, courts afford deference according to the persuasiveness of an agency’s reasoning, respecting professional judgment when it is supported by a well‑articulated justification. This form of deference enhances rational judicial decisions and promotes “regulatory excellence” by requiring agencies to convince through logic and evidence rather than rely on sheer authority. As the article notes, treating agency reasoning as persuasive “serves as a superior substitute for deference and realizes democratic accountability.”
The Hidden Pitfall: Algorithmic Deference
While courts now scrutinize how agencies interpret the law, administrative law lacks a framework for monitoring the implicit policy choices regulators make when they implement those laws using algorithms or AI. As AI tools increasingly infiltrate agency decision‑making, problems such as hallucinations and sycophancy render these outputs neither convincing nor accurate. Danielle Citron’s empirical work shows that “encoded policy receives more than Chevron‑level deference,” meaning courts may inadvertently defer to opaque algorithmic decisions. Because technical biases and policy choices are hidden within code, they evade meaningful public or judicial scrutiny. The article warns that “we do not need agencies that merely command; we need agencies that can convince,” emphasizing that accountability must rest on human judgment, not unexamined computer outputs.
Normative Justification for Persuasion in AI Governance
Adopting a persuasion‑based approach entails uncertainty and higher deliberative costs, but these costs are normatively justified in fields like AI governance. Epistemic uncertainty and systemic risks demand that the reasoning behind decisions be transparent and open to challenge. By requiring agencies to justify their interpretations, the system invites iterative feedback from state tort litigation, academia, industry, and the public. This feedback loop helps calibrate regulations to the fast‑evolving AI landscape, reducing the danger of rigid, misaligned rules that stifle innovation or fail to protect citizens.
Conclusion: Toward a Convincing, Accountable Administrative State
The Loper Bright decision ended the era of automatic Chevron deference, yet the underlying tension between federal authority and state tort power persists—especially in the governance of artificial intelligence. A regulatory regime that treats agency interpretations as persuasive, grounded in transparent reasoning, preserves the information‑producing function of state litigation while providing a necessary baseline of federal standards. Moreover, vigilance against algorithmic deference is essential to ensure that encoded policy does not escape scrutiny. Ultimately, effective AI oversight depends not on agencies that command through authority, but on those that can convince through reasoned, accountable judgment.