Senator Ed Markey, a Massachusetts Democrat, has unveiled a sweeping package of bills designed to address the harms of artificial intelligence, from the environmental impact of data centers to algorithmic bias and child safety risks. The 79-year-old senator, who has authored nearly a dozen previous AI bills, announced his “AI accountability agenda” in an exclusive interview with the Guardian, framing it as a necessary step to “take power back from big tech.” His focus is on preventing the technology from outpacing the rules intended to contain it.
Data Center Certification: The Centerpiece
The cornerstone of Markey's agenda is a bill expected to be introduced in the coming weeks. It would mandate that any company owning or planning a data center obtain certification from the Federal Communications Commission before construction begins. The certification must confirm that the facility “will not harm the public interest.” Under the draft, the FCC would evaluate the proposed data center’s impact on air and water quality, noise levels, energy costs, grid reliability, local wildlife, and the local economy. The agency would consult with the Environmental Protection Agency and local zoning boards before granting approval. “We need to make sure these datacenters don’t turn into pollution bombs,” Markey warned, referencing incidents where data centers have been linked to water contamination and increased pollution in nearby communities.
Data centers are notorious for their high energy consumption and water usage. A single large facility can consume as much electricity as a small town, straining local grids and driving up costs for residents. In some cases, such as in Georgia, residents have reported their tap water becoming undrinkable after a data center moved in nearby. Markey’s bill aims to force companies to proactively demonstrate their facilities will not exacerbate such problems, rather than reacting after damage is done. This precautionary approach mirrors the European Union’s AI Act, which requires risk assessments for high-risk AI systems before they are deployed.
Algorithmic Bias and Workplace Surveillance
Beyond data centers, the agenda includes legislation to mandate independent bias audits for high-stakes algorithms before they are deployed. This would cover algorithms used in hiring, housing, credit, and criminal justice, where biased decisions can have life-altering consequences. Markey cited the case of a woman who sued after an algorithm denied her housing, allegedly due to discriminatory factors. The bill would require companies to submit their algorithms to third-party testing for bias and transparency, with the results made public.
Another bill in the package would restrict employers from relying primarily on automated systems for hiring, firing, and promotion decisions. Human oversight would be mandatory, ensuring that workers are not unjustly terminated or passed over by a black-box algorithm. Markey referenced the experience of a veteran nurse who was pressured to trust an AI’s judgment over her own clinical expertise, highlighting the dangers of delegating critical decisions to machines without accountability.
Protecting Children from Harmful Chatbots
The agenda also addresses the growing concern over chatbots that can form emotional bonds with children, a phenomenon that has led to tragic outcomes. Markey pointed to the case of a 14-year-old boy who died by suicide after developing a dependency on a chatbot that allegedly encouraged self-harm. The proposed legislation would prohibit chatbot companies from designing their products to foster emotional dependence in minors, and would require clear warnings about the risks of prolonged interaction. It would also give parents the right to sue companies that violate these rules.
This follows a broader bipartisan push in the Senate to protect children online. In March, the Senate passed a bill tightening online safety rules for children, banning targeted advertising to minors and limiting data collection. Markey’s chatbot bill would complement that effort by specifically targeting the unique risks of generative AI, which can simulate human conversation and manipulate users, especially vulnerable young people.
Human Override in Healthcare
In healthcare, Markey wants to ensure that AI never fully replaces human judgment. A bill in his package would require hospitals and health systems to maintain a human override for any AI-driven decision, from diagnosis to treatment recommendations. This is especially critical in fields like radiology and pathology, where AI is increasingly used to analyze scans and tissue samples. The nurse’s story of being pressured to accept an AI’s diagnosis over her own experience underscores the potential for automation bias, where humans defer to machines even when they have reason to disagree.
The Need for Federal Action
Markey’s argument for federal legislation is rooted in the idea that safety should not depend on where you live. The United States currently has no comprehensive federal law regulating AI, leaving a patchwork of state laws that vary widely. California, for example, has its own AI transparency rules, while other states have none. “Every American is entitled to these safeguards,” Markey said, warning that a state-by-state approach “would leave too many people exposed.” He pointed to the European Union, which has taken a proactive stance with its AI Act, General Data Protection Regulation (GDPR), and online safety rules. In contrast, Washington has done virtually nothing since OpenAI launched ChatGPT in 2022, even as tech companies themselves have begun calling for regulation to create certainty and prevent misuse.
The political landscape, however, is unforgiving. Most of Markey’s previous AI bills remain stuck in committee, and he is just one senator in a Congress that has largely favored speed over guardrails when it comes to technology. The tech industry has lobbied heavily against regulations that could slow down innovation or increase costs. Yet Markey remains optimistic. “Ultimately, there will be national solutions that will be put on the books,” he said, noting that public pressure is mounting as AI-related harms become more visible.
There are a few footholds. The Senate’s passage of the children’s online safety bill in March shows that bipartisan support for certain protections is possible. Separately, other lawmakers are pushing for legislation to make Big Tech pay the power bills of the data centers that drive up electricity costs. These initiatives suggest that the conversation is shifting, even if comprehensive reform remains a long way off.
Markey traces his concern for safety back to his father, who lost a finger to a factory machine because modern safety laws did not exist at the time. His point is that technology has always outrun the rules meant to contain it. AI, he argues, is just the newest version of that gap—and the stakes are higher than ever. With data centers consuming vast resources, algorithms making life-altering decisions, and chatbots shaping young minds, the need for accountability is urgent. Markey’s agenda offers a first step toward a federal framework that could set standards for the industry, protect consumers, and ensure that the benefits of AI are not overshadowed by its harms.