Senator Markey proposes federal law to regulate AI harms

What happened?

U.S. Senator Edward J. Markey (Massachusetts) on Friday introduced a comprehensive legislative package called the AI Accountability Agenda. According to The Next Web, the proposal includes several bills covering everything from the water and energy footprint of data centers to workplace surveillance, biased algorithms, and chatbots interacting with minors. Markey seeks to turn into federal law what has so far been a patchwork of state regulations. The agenda comprises at least six bills, some of which had been introduced in previous sessions without success, but are now grouped under a common umbrella to increase visibility and political pressure.

Why is it important?

The current regulatory fragmentation creates uncertainty for tech companies and leaves protection gaps for citizens. A federal law would unify criteria, establish minimum standards, and create clear accountability mechanisms. It also addresses emerging issues like AI's environmental impact, which has received little legislative attention so far. The political moment is key: generative AI has skyrocketed demand for computing and energy, while complaints about algorithmic bias in hiring, credit, and criminal justice grow. According to the International Energy Agency (IEA), data center electricity consumption could double by 2026, reaching 4% of global total. In the U.S., AI data centers alone consumed about 50 TWh in 2025, equivalent to the residential consumption of the entire state of Massachusetts. Additionally, water use for cooling is critical: a University of California study estimates that training GPT-3 consumed 700,000 liters of water, and larger models like GPT-4 could require up to 5 million liters.

Consequences for the sector

• Data centers: Companies would have to report water and electricity consumption and could face limits in water-stressed areas. The proposed law, called the Data Center Accountability and Transparency Act, would require quarterly reports to the Environmental Protection Agency (EPA) and set fines of up to $100,000 for non-compliance. Companies like Google and Microsoft have already announced sustainability goals, but without a legal framework, these are voluntary and non-binding.
• Workplace surveillance: Continuous monitoring without explicit consent would be banned, affecting remote work platforms and productivity software. The Stop Spying Bosses Act would require employers to notify workers about any surveillance systems and obtain prior authorization. This would directly impact companies like Amazon, which uses cameras and sensors in its warehouses, and platforms like Upwork or Fiverr that track freelancer activity.
• Biased algorithms: It would be mandatory to audit AI models used in high-impact decisions (housing, employment, credit) and correct demonstrated biases. The Algorithmic Accountability Act would require independent impact assessments for any automated system making significant decisions. A 2024 FTC study found that 40% of AI-based hiring tools showed racial or gender bias, and only 15% of companies conducted external audits.
• Child protection: Chatbots and virtual assistants aimed at minors would have to meet stricter safety and privacy standards. The Children and Teens’ Online Privacy Protection Act (COPPA 2.0) would ban targeted advertising to minors and require chatbots to have mechanisms to detect and report harmful content. According to Common Sense Media, 60% of teens have interacted with chatbots, and 30% have unknowingly shared sensitive personal information.

What should readers know?

Markey's agenda still must go through committees and votes; its approval is not imminent. However, it marks a shift in tone in Washington toward more interventionist AI regulation. Companies like OpenAI, Google, and Microsoft have expressed willingness to collaborate but warn that overly rigid rules could stifle innovation. For now, the debate centers on balancing accountability and competitiveness. Notably, Markey is known for his critical stance toward big tech: he co-authored the 1998 Children's Online Privacy Protection Act (COPPA) and has led investigations into anticompetitive platform practices. His agenda is backed by organizations like the ACLU, Electronic Frontier Foundation, and the AFL-CIO union, but faces opposition from business groups like the U.S. Chamber of Commerce, which argues regulations could cost billions in compliance and delay AI adoption in sectors like healthcare and logistics.

“AI cannot be a wild west without rules. We need accountability from design to deployment,” Markey said at the presentation.

Context and comparisons

The initiative echoes the European Union's AI Act, which classifies systems by risk level. However, Markey's approach is more specific on environmental and labor issues. Unlike previous proposals, such as the 2022 Fair Algorithms Act, this one explicitly includes data centers as critical infrastructure subject to regulation. It also differs from President Biden's October 2023 Executive Order on AI, which focused on national security and fairness but lacked mandatory enforcement mechanisms. Markey's agenda, if passed, would be the first comprehensive federal AI law in the U.S., comparable to the GDPR in Europe in scope. However, the legislative path is uncertain: in the current Congress, with a slim Republican majority in the House and a divided Senate, the agenda is likely to face significant amendments or even a blockade. Markey has indicated he is willing to negotiate but insists that the core principles of transparency and accountability are non-negotiable.