Desiree Sainthrope is a legal heavyweight navigating the intersection of trade agreements and emerging tech. As a recognized authority on compliance and intellectual property, she offers a sharp perspective on the friction between state transparency laws and federal oversight. Today, she discusses the high-stakes battle over the RAISE Act and the future of American AI governance.
State lawmakers are increasingly concerned that federal preemption could dismantle local transparency laws like the RAISE Act. How would a federal block on state-level reporting requirements impact public safety, and what specific risks might go unnoticed if companies aren’t forced to disclose model failures?
When federal preemption sweeps away state reporting, we lose the early warning systems designed to protect the public from algorithmic fallout. If laws like the RAISE Act are neutralized, companies can keep catastrophic risks tucked away in private servers, away from the prying eyes of state regulators who are more agile than Congress. This “paper over” approach prevents the public from knowing when powerful AI systems fail or what specific dangers they pose to our daily infrastructure. We risk a scenario where a massive failure occurs, but because disclosure wasn’t mandatory, the danger remains hidden until it’s far too late to mitigate.
Negotiations in Congress have fluctuated between mandatory reporting and voluntary industry standards for top AI developers. What are the practical trade-offs of a voluntary regime, and how can the government effectively slow the release of dangerous models without stifling innovation?
A voluntary regime essentially asks the public to trust corporations whose primary drive is profit rather than safety, which is a fundamentally risky proposition. In current negotiations, lawmakers are struggling to decide if the government should have the authority to slow or block the release of cutting-edge models it deems dangerous. To protect citizens, we need a legal framework that creates clear, mandatory rules of the road rather than leaving safety as an optional checkbox for developers. This balance is critical to ensure that rapid innovation doesn’t come at the cost of national security or public welfare.
Massive financial investments from pro-AI super PACs are now influencing political campaigns focused on tech regulation. In what ways does this level of industry spending alter the legislative process, and what steps can be taken to ensure safety standards prioritize public welfare over corporate profits?
The infusion of capital is significant, with the “Leading the Future” network spending more than $1 million to target specific lawmakers who support strict AI oversight through aggressive anti-reform ads. This massive spending creates a high-pressure environment where industry giants can effectively “aid and abet” attempts to set rules that only benefit their own power and profits. It shifts the legislative focus away from public safety, making it much harder for lawmakers to implement a multi-faceted approach to AI’s potential downsides. We must ensure that the voices of concerned citizens aren’t drowned out by the metallic clink of corporate lobbying dollars.
Proponents of state action argue that a multi-faceted, local approach is necessary when federal progress stalls. How do conflicting state and federal transparency requirements complicate operations for cutting-edge AI developers, and is there a middle ground that preserves state oversight while maintaining a unified national standard?
Developers argue that a “patchwork” of rules from New York to Massachusetts complicates global product releases and slows down their technical workflows. However, state leaders see this as a necessary safeguard when the federal government fails to implement a comprehensive approach to address AI risks. A middle ground would require a federal floor that provides a baseline of safety without blocking states from acting when they see immediate local threats. Without this, we are essentially letting the industry dictate its own level of accountability while state hands remain tied.
Public transparency is often cited as a critical tool for identifying catastrophic risks in powerful AI systems. Can you walk through a step-by-step scenario where local reporting would prevent a large-scale failure, and what metrics should be used to determine if an AI model is too dangerous for release?
Imagine a powerful AI system used in public services that starts to exhibit catastrophic failure traits during a localized rollout. Under a state transparency law, the developer would be legally bound to report this failure immediately, allowing state engineers to intervene before the glitch cascades into a full-blown emergency. Metrics for these releases should include specific, public reporting on model failures and the identified risks to public safety that occur during stress tests. This transparency ensures that the public isn’t the last to know when a system becomes a liability rather than an asset.
What is your forecast for AI regulation?
I expect the tug-of-war between state sovereignty and federal preemption to intensify as more states follow New York’s lead in demanding transparency. We will likely see more aggressive lobbying and million-dollar campaigns aimed at silencing proponents of mandatory reporting as the stakes for these tech giants continue to rise. Ultimately, the success of AI regulation will depend on whether we prioritize public transparency over the unchecked growth of corporate power. The era of secret, high-risk AI experimentation is coming to an end, one way or another.
