Former US representative Brad Carson of Oklahoma is president of Americans for Responsible Innovation.<\/p>\n
Representative Lori Trahan of Massachusetts is right that Congress should pass national safeguards for artificial intelligence. But that work cannot begin by taking away the one source of AI accountability that has actually moved while Congress has stalled: the states, including Massachusetts.<\/p>\n
A draft bill<\/a> filed this month by Trahan, a Democrat, and Republican Representative Jay Obernolte of California threatens to freeze state regulation of companies<\/a> developing AI for the next three years. Specifically, it would block states from enforcing laws that regulate AI model development and prerelease safety decisions on the grounds that those regulations impact interstate commerce. That\u2019s the stage where training data is chosen, models are tested, risks are measured, and release decisions are made. In exchange for that ban on state regulation, the bill proposes a set of transparency measures and government programs without the teeth to hold AI companies accountable for unsafe practices.<\/p>\n This is not an abstract fight about federalism. Every week, another story surfaces of a child steered<\/a> toward self-harm and violence by unsafe chatbots. AI models are supporting discriminatory hiring<\/a> decisions. All the while, Big Tech is lobbying<\/a> to kill regulation and oversight to maintain its \u201cmove fast and break things\u201d ethos. While Congress has stalled, lawmakers in Massachusetts and other states have moved quickly<\/a> to put basic guardrails in place.<\/p>\n There is no question that Congress should act on AI. National AI safeguards are needed for a technology that crosses state lines and impacts all Americans. But after more than a decade of federal inaction on social media harms, state lawmakers are the first responders<\/a> in the face of emerging impacts from rapidly advancing technology. Pausing a broad range of state AI laws for the next three years would put families and workers at enormous risk.<\/p>\n While Trahan\u2019s proposal asks AI companies to disclose how they test new AI models for safety, nothing in it prohibits them from using subpar standards to evaluate their models. While the draft bill funds federal research into what might constitute safe AI, nothing in it prevents an AI company from releasing a model that fails to meet child-, worker-, or consumer-safety standards.<\/p>\n That is probably the reason why tech industry lobbyists<\/a> have applauded the legislation while a broad coalition of labor unions<\/a>, child-safety advocates<\/a>, civil liberties groups<\/a>, and grass-roots organizations<\/a> have raised the alarm.<\/p>\n Their concerns are well-founded. Banning state laws on AI development would stop states from, for example, requiring that AI companies scrub child abuse content from training data. The ban could also prevent states from requiring AI companies to disclose whether models are trained on copyrighted artistic content. Nationwide, the bill would freeze the progress we\u2019ve seen from states in demanding transparency on basic safety processes from AI companies.<\/p>\n To be clear, Trahan\u2019s legislation does not ban every state AI law. Its sponsors are right that the draft legislation preserves laws of general applicability and many state rules that apply after deployment. But that carveout is not enough. To reduce AI harms before they occur, regulators need authority before deployment, while model design, training data, safety testing, and release decisions are still being made.<\/p>\n A three-year freeze on state AI laws will likely also be weaponized by tech companies, who will sue in court to expand the definition of what state laws are prohibited by Trahan\u2019s proposal. Just as Big Tech weaponized laws like Section 230<\/a> to shield themselves from lawsuits over social media harms, expect Silicon Valley to use this legislation to sue states over a long list of regulations, from deepfake bills to data center restrictions.<\/p>\n Tragically, while Congress considers stalling AI legislation in the states, AI risks are advancing rapidly.<\/p>\n According to a recent study<\/a>, 86 percent of students are now using AI, with consequences ranging from students feeling less connected to their teachers to a decrease in social connections. In the workplace, AI has become one of the leading causes companies cite for mass layoffs<\/a>. Surveys show<\/a> that 8 in 10 people in creative careers are worried about their style, voice, and likeness being stolen by AI.<\/p>\n Instead of offering a draft bill that sets a federal ceiling on AI regulation and prevents state lawmakers from addressing emerging AI harms, Congress should set a floor: requiring tech companies to meet real standards that protect American communities without freezing state authorities.<\/p>\n