We Need NHPA Reform to Enable AI

The Free State Foundation has written many times about how difficulty obtaining necessary government permits and rights-of-way adds unwarranted costs and\or delays to deploying broadband. Although most permitting problems are at the state and local levels, the federal government also requires costly studies for permission to use federal property. The two principal statutes are the National Environmental Policy Act (NEPA) and the National Historic Preservation Act (NHPA). On Tuesday another voice weighed in. “The Permitting Window is Closing,” by Tahra Hoops of The Rebuild urges Democrats to make NHPA reform a priority and negotiate a bipartisan agreement. However, because of the problems associated with future energy supplies, the call for support should be directed to all sides.

NHPA is one of two significant federal statutes that govern activity on federal lands. Although good data on costs and delays does not exist, a paper from earlier this year estimated that, absent reform, the cost of both statutes for just outdoor wireless facilities would be $2.2 billion over the next decade. Overall economic harm would total $7.5 billion. This is equal to 18 percent of the Broadband Equity Access and Deployment program. The study found that: “the current permitting process adds unnecessary costs for wireless infrastructure and service providers and delays the deployment of higher-capacity networks and innovative services in the United States.”

 

 

NHPA requires federal agencies to evaluate the impact of federal undertakings on historic property. Section 106 requires agencies to identify historical properties that may be affected by a proposed undertaking and assess whether the action would affect those properties. Broadband deployments are considered undertakings and therefore require a more detailed study.

Ms. Hoop’s analysis looks at NHPA primarily through its effect on electricity markets. A combination of growing demand and supply constraints has caused energy prices to rise sharply over the last few years. Ms. Hoops points to data from the Energy Information Administration showing that the cost of electricity rose 42 percent over the last five years. On average, households spent $110 more in 2025 than in 2024. This causes a slew of problems. One is that the backlog in rights-of-way approvals prevents a great deal of clean energy from connecting to the electric grid. To make things worse, the reduction in supply causes power plants to ramp up their use of coal and oil.

Second, U.S. policy prioritizes goals that, if pursued, will require tremendous amounts of new electricity. A key aspect of U.S. climate policy is the electrification of everything from vehicles and air conditioning to energy storage. Even if the electricity to power these machines is coal, electrification will still require large increases in supply. Then there is artificial intelligence (AI), which requires incredible amounts of electricity to power data farms and run the information networks needed to connect AI to the rest of the economy. Without a modern sophisticated collection of networks, including connections to the power grid, AI will not amount to much. Under current policy, AI capacity and needs are forecast to rise rapidly, so any policy that makes it easier to build and connect power plants on federal lands helps advance AI.

Permitting reform can be accomplished. Utah’s State Historic Preservation Office digitized decades of paper records into a GIS system in 2017. Today 98 percent of its reviews clear within seven days, saving the state roughly $250,000 per year. The Bipartisan Policy Center recently published a paper on possible permitting reforms for the NHPA. Commonly mentioned changes include tighter time deadlines. reducing the number of properties covered by the Act, limits on judicial review, and digitizing records.

NHPA reform will have lots of benefits. The environmental ones are not always clear but as Ms. Hoops says: “[a] technology-neutral permitting overhaul is a net-clean policy by simple arithmetic. Speeding up everything speeds up clean energy most, because clean energy is what’s waiting in line.”

A reduction in federal permitting time and costs will speed up deployment projects and reduce the cost of spreading broadband to all parts of the country. Finally, shortening the permitting process will reduce the cost and time associated with finding new sources of electricity and connecting them to the grid. That has a tremendous effect on the growth of AI. As stated before, this is one of the key constraints to the buildout of the new information networks needed to convey, compute, and control the massive amounts of data needed to support AI.

Congress is currently considering a number of reform bills including the Historic Preservation Fund Reauthorization Act (H.R 3416). Last session a bipartisan bill passed committee but members were unable to seal the deal. The environmental, power, and scientific benefits remain great. Let’s hope they do this year.

The Growing Realization That AI Needs Modernized High-Capacity Information Networks

It is becoming increasingly difficult to avoid discussions about artificial intelligence (AI). There is a growing consensus that AI broadly defined will have a major impact on virtually every sector of the economy. A growing number of experts believe that the impact will not stop there. Like railroads, electricity, telephones, steam engines, and the Internet, AI’s impact is expected to be large enough to affect the way we live. But, as Free State Foundation scholars have written, this will require massive investment to modernize the nation's information networks and to keep them robust.

Most of the attention on AI is focused on building models that will consume massive amounts of data and compute complex problems that are increasingly beyond the ability of humans to solve. A lesser concern has been the energy and, to a smaller extent, water networks required because of their role as major inputs into data centers and power plants. However, AI will also have a large impact on other networks. One of the most important will be the nation's information networks. 

 

In the past, most of the focus on networks has been on extending broadband to all people, a job that is nearing completion as broadband availability becomes ubiquitous. Over the past year or so an increasing number of leaders and organizations have begun to point out the strong interdependency of AI and the information networks. The result is a growing realization that the U.S. needs to devote an enormous amount of investment to networks that are larger, faster, and more self-aware than the existing infrastructure.

As stated above, there is already wide recognition of the strong linkage between AI and power supplies. For instance, Satya Nadella, CEO of Microsoft, has stated that the problem in the AI industry is not an excess supply of computation power, but rather a lack of power to accommodate all those CPUs. Jensen Huang, NVIDIA's CEO, has stressed that the U.S. is vulnerable because of its deficient energy supply. Finally, a report by the Center for Strategic and International Studies finds that the U.S. electricity sector is struggling to meet growing demand while maintaining low costs, improving system reliance, and reducing emissions.

Recently authoritative voices have expressed some of the same "supply-based" concerns about the information networks. These networks must convey, compute, and control massive amounts of data to massive amounts of computing power and back. Börje Ekholm, President and CEO of Ericsson, explained that “[a]s artificial intelligence (AI) moves beyond data centres into real-world applications like robotics, autonomous systems and extended reality, it depends on high-performance 5G today and 6G tomorrow.” John Saw, T-Mobile's President of Technology and Chief Technology Officer, believes that “6G to us is more than just an ‘XG.’ We think it's the foundation for an AI-native future that distributes intelligence across devices, the edge, and the cloud.” Finally, Ajit Pai, President and CEO of CTIA stated that: “AI without a strong wireless network is like a new car without a road.”

Others share these concerns. An informative report from the Fiber Broadband Association argues that “[t]wo historic trends are unfolding at the same time: the nationwide deployment of fiber broadband infrastructure and the rapid buildout of the infrastructure required to support artificial intelligence, quantum networking, and other emerging applications.” FBA's report says: “AI workloads require high-capacity east-west traffic within and between data centers. They require low-latency pathways between inference platforms and end users. They require resilient interconnection among geographically distributed facilities.” The report argues that the current grid is evolving from a centralized system into a highly distributed network incorporating renewable energy resources, battery storage systems, distributed generation, microgrids, and intelligent controls. Managing this complexity requires real-time visibility and coordination.

A recent CTIA report argues that AI requires networks to move data, coordinate real-time decisions, and interact with the physical world. In turn, wireless networks rely on AI to manage the surging complexity and record traffic driven by AI’s own insatiable data demands:”

[I]t is now clear that AI traffic will strain existing wireless networks before the decade is out with huge new data needs, entirely new traffic patterns, and novel demands on wireless networks to do more than simply carry traffic….

[I]t will also require emerging 6G networks to be AI-native from the ground up with embedded intelligence to dynamically allocate spectrum, anticipate congestion, sense the physical environment, coordinate edge-compute workloads, and secure devices—all at machine speed.

Two final points. To maximize AI’s performance, the most important parts of the networks have to work differently than current networks. They will feature more east-west flows, lower latency, higher uploading speeds, and the ability to operate independently of humans. Second, the networks will have to be closely integrated into those of other industries, including healthcare, transportation, government services, and (of course) electricity.

In fact, building out modernized networks will require both fiber and wireless technology, as well as a lot of other inputs. Success will require massive investments in these modernized networks, most of which will come from the private sector. Given the large economic and security implications, public policy should concentrate on creating favorable conditions for private sector investment and working with allies to develop common standards and protections.

 

AI Plus 6G: A Convergence of Mutual Necessity

In conjunction with its annual summit on May 6, CTIA issued a new report devoted to the ongoing merger of wireless technology and AI: Wireless & AI: Driving the Future of Innovation. The report points to the growing co-dependency between the data networks (wired and wireless) and artificial intelligence (AI), a technological advancement that promises to create great value. In doing so, the report discusses several main points that are likely to drive the future of both AI and broadband.

The first principle, which I discussed in an FSF Perspectives earlier this year, is that AI and networks are increasingly interdependent. As the CITA report mentions: “AI requires wireless networks to move data, help coordinate real-time decisions and operate effectively with the physical world. In turn, wireless networks rely on AI to manage the surging complexity and record traffic driven by AI’s own insatiable data demands.

AI without data networks is useless. AI needs the models to download vast amounts of data of all types (code, text, sound, and visual), transport them to any location, process the information, transmit the analysis, and increasingly, act on the results itself without human intervention, a phenomenon the report refers to as "Physical AI." AI will operate across three layers; devices, edge, and the cloud, depending on the speed, complexity, and cost of the task.

Second, the dependence is two-way. While AI is heavily dependent on communications networks to maximize its use of data, the rapid increase in AI’s use of networks demands that the networks use AI to maximize their capacity in the form of transmission capacity, latency, and compute power. Just as the global power of AI depends on networks’ capacity, that capacity must use AI to expand in order to meet these increased needs. 6G networks will respond dynamically to the conditions and demands facing them. Already some are predicting that by the end of the decade one-third of AI traffic needs could go unmet. Accenture estimates that this could reduce potential GDP by $1.4 trillion.

The third point is that the physical merger is already occurring. 6G networks will increasingly connect with a wide variety of machines and sensors. Unlike existing networks, AI will require massive amounts of data to be uploaded into communications networks for analysis. AI traffic is expected to power 75 percent of smartphones within two years. Already, AI traffic is growing three times faster than overall traffic and is expected to account for 30 percent of networks’ traffic by 2034. Finally, 6G networks are expected to increase energy efficiency by 30 percent.

These advancements impact economic growth and national security. AI and wireless are the two top sources of infrastructure investment in the current economy, offsetting some of the uncertainty caused by higher consumer prices and international conflict. The significant dual use capability of AI and network advancements creates significant security implications and places a premium on intelligent and timely regulation. The world’s complexity increasingly cannot be resolved by humans manually reacting to data flows that move faster than human reaction time.

The report focuses on two major areas for policy reform. The first is the allocation of large contiguous blocks of licensed spectrum. The value of spectrum has grown rapidly due both to new uses and the increased capacity of existing uses. As a result, the allocation of spectrum is attracting increased demands from government, industry, and consumer use. The sooner Congress and regulators can develop methods for allocating spectrum to its most valuable uses, the better. The report points to next year’s World Radiocommunication Conference as an important milestone for developing a Western response to international policy.

Finally, permitting reform will also play a large role in determining the pace of innovation. Companies frequently need government approval at the federal, state, or local level before they can start building out Internet infrastructure, whether in the form of home broadband, data centers, or transmission lines. At the state and local levels it is not uncommon for agency officials to demand high fees or costly extraneous requirements as a condition to start construction. While some progress has been made at the federal level, state and local entities still impose significant delays. Both the FCC and Congress need to continue to remove impediments and implement meaningful permitting reform.

The ongoing merger of the communications networks (including eventually 6G) and AI will have vast implications for society. Machines will gather more information, transmit it widely, analyze any correlations within it, and act on the results. This will dramatically expand the information available to humans. The challenge is to use it wisely and for the benefit of all. AI without reliable, secure, high-capacity communications networks is useless, but the networks without AI will collapse.

Maryland Doesn't Need to Stop Dynamic Pricing

Developments in artificial intelligence continue to raise alarm among the public and lawmakers. Among the many concerns cited about artificial intelligence and automation is dynamic pricing. To this end, Maryland Governor Wes Moore signed legislation last week banning grocery stores and third-party delivery services from using individual shopper data to increase prices "dynamically."

Under dynamic pricing, sellers may use data about shopping behavior to automate and continuously adjust their prices. Under individualized dynamic pricing – sometimes called surveillance pricing in pejorative terms – businesses set different prices for different consumers by charging more to shoppers who appear willing to pay a premium or offering lower prices to customers who might not otherwise buy. Other types of dynamic pricing may include shifting prices at different times of day based on changes in demand or competitive conditions.

The underlying logic of dynamic pricing is straightforward: businesses have always tried to match price to demand, and data-driven tools make doing so easier.

Maryland’s bill drew public support, reflecting broader concern with companies exploring individualized pricing, especially on food and housing as basic needs. Critics frame these practices as predatory: corporations using shadowy algorithms to target and extract as much money as possible from individual shoppers.

However, the alarm reflects a misconception regarding what data collection and algorithmic pricing can actually accomplish. Even the most sophisticated artificial intelligence uses incomplete information and thus imperfect predictions – the same reason why centrally planned economies with government-dictated prices are so inefficient. Consumer preferences change with income, season, family circumstances, competing options, and other infinite variables that are impossible to capture in a dataset. The premise that an algorithm can reliably identify each shopper's maximum willingness to pay overstates the role that data and algorithms play in society.

Dynamic pricing is also already a routine feature of commerce. Airlines adjust fares continuously based on demand, booking patterns, and seat availability. That's why the person sitting next to you on a plane likely paid a different price than you paid for her ticket. Bars and restaurants offer happy hour pricing. Retailers run flash sales, time-limited promotions, and personalized discounts. Even Maryland’s own law acknowledges this reality with its numerous exemptions and clarifications for longstanding practices – promotional pricing, loyalty program discounts, and other temporary price reductions.

Moreover, the alarm over dynamic prices overlooks the consumer benefits. A grocer or other business that makes more sales has more room to keep overall prices low, and dynamic individualized prices can be what closes a sale that otherwise would not have happened. This means that people can buy things that otherwise wouldn’t have fit in their budgets.

Maryland’s law purports to address a public concern by conflating a common business practice with a supposedly harmful predatory practice and without acknowledging the consumer benefits. Maryland should indeed tackle deceptive trade practices in grocery stores and elsewhere, but states should not ban technology before actual harms to consumers materialize. Regulating against possible harms has its consequences – shoppers forgo benefits that they never even see.

Sanders' AI Bill Is a Red Herring and Blackburn's Has Problems

With the White House calling for a national AI framework to end the patchwork of state regulation, two notable proposed pieces of federal legislation have emerged. And the one getting less attention at the moment is the one that matters more.

Senator Marsha Blackburn (R-TN) released a discussion draft of the TRUMP AMERICA AI Act (you read that right, The Republic Unifying Meritocratic Performance Advancing Machine Intelligence by Eliminating Regulatory Interstate Chaos Across American Industry Act) on March 18, 2026, a 291-page federal framework developed in response to the Trump administration's call for a national AI policy. Senator Bernie Sanders (D-VT), joined by Representative Alexandria Ocasio-Cortez (D-NY), introduced the 13-page Artificial Intelligence Data Center Moratorium Act on March 25, 2026. It would halt data center construction until Congress enacts legislation to ensure: that future AI products are "safe and effective"; that AI does “not threaten the health and well-being of working families”; and that AI does not displace jobs. The two AI bills are not comparable in scope or consequence.

The Sanders moratorium bill has received the most mainstream coverage, possibly in part because it is the only one of the two to be formally introduced. But it’s easy to see why all the fuss. The moratorium bill takes advantage of anxieties that translate directly into headlines: job displacement, strain on the power grid, and industrial construction in people's backyards. While these concerns affect real people, the bill's moratorium is ill-conceived and would be harmful. Pausing data center construction pending new AI legislation would be a significant brake on American AI infrastructure at precisely the moment the Trump administration is pushing to accelerate it and would let foreign competitors move ahead.

But Sanders’ moratorium bill is almost certainly a political statement about AI as a threat rather than a realistic proposal. It is unlikely to gain serious legislative traction, and its primary practical effect may be to divert attention from more consequential legislation.

The Blackburn bill is one piece of potentially more consequential legislation. As a proposed comprehensive federal AI framework, it is more technically complex and far-reaching than the moratorium bill. Yet it has received a fraction of the coverage. Other think tanks including the Competitive Enterprise Institute and the Cato Institute have explained how the bill would impose heavy-handed regulation across the AI ecosystem.

Some aspects of Senator Blackburn’s bill that may be problematic and require close attention include: a full-on repeal of Section 230 of the Communications Act of 1934; imposing “duty of care” on chatbot developers; holding AI developers liable for harms beyond existing laws on fair and deceptive practices; requiring federal contracts to use "unbiased" large language models; creating a Department of Energy testing program for adverse incidents in AI systems; and directing DOE to develop certification procedures, licensing requirements, and broad regulatory oversight.

I wrote last week that the federal AI framework needs a light-handed approach grounded in free market competition. I explained that “robust competition among American companies is the precondition for national competitiveness” and consumer satisfaction. While established developers may fare fine under such a burdensome scheme, their products would fall behind other nations not facing such operating and compliance costs. And startups and emerging competitors would fare even worse.

The Sanders moratorium deserves the criticism it has received. But the current Blackburn bill has problematic provisions that deserve scrutiny it has not yet gotten.

White House to Congress: Fix the AI Patchwork

The Trump Administration issued seven AI policy recommendations for Congress on Friday, March 20, 2026, including one for preemption, asking Congress to make sure state legislatures don’t get in the way of AI innovation (Recommendation VII). 

This recommendation is exactly what the moment calls for. Last week, I wrote a FSF Blog post about just this issue. After describing the extraordinarily wide range and volume of AI bills moving through statehouses across the country, I wrote: “What the nation really needs is an overarching federal framework that avoids ex ante heavy-handed regulation and that supplants the growing patchwork of state laws.” The White House has now said the same thing. 

Recommendation VII reads: “Congress should preempt state AI laws that impose undue burdens to ensure a minimally burdensome national standard consistent with these recommendations, not fifty discordant ones.” It clarifies the distinction between federal and state domains of AI regulation. The federal government is better positioned to “supporting innovation” because AI is “an interstate phenomenon” that is part of the “national strategy to achieve global AI dominance.” Absent preepmtion, states may otherwise “unduly burden Americans’ use of AI.” State governments are positioned to regulate AI as it pertains to issues specific to their state such as consumer protection, zoning, law enforcement, and public education.

Here are a few additional details encouraging innovation among the White House’s six other recommendations: “lead the world in AI by removing barriers to innovation” (Recommendation V); “not create any new federal rulemaking body to regulate AI” (Recommendation V); and “streamline federal permitting for AI infrastructure construction and operation” (Recommendation II); The White House also recommends preventing censorship and protecting free speech (Recommendation IV).

Noticeably absent from the seven recommendations, however, is an explicit acknowledgment that free market competition is both a means of achieving the White House’s ambitions and an essential benefit to American consumers. Recommendation VII frames preemption in terms of competing with other nations but overlooks a foundational point: robust competition among American companies is the precondition for national competitiveness. A truly pro-innovation framework would make free market competition an explicit objective in recognition that this helps ensure that the best products and services are made available to consumers at the lowest prices. 

Now, Congress needs to follow through on the White House’s recommendation for preemption. And with a strong commitment to fostering market competition, Congress and existing federal agencies have the opportunity to get AI regulation right.

State Lawmakers Are Not Waiting for Washington to Regulate AI

With annual legislative sessions beginning to wind down, lawmakers in 44 states and D.C. have introduced over 800 bills related to artificial intelligence during the 2026 session so far, according to the National Council of State Legislature’s AI bill tracker. This is a remarkable volume of regulatory interest from lawmakers who have not had much of a chance to understand any possible related market failures or to study the costs and benefits of regulations for a new technology that has only recently entered mainstream use. This regulatory interest represents continued momentum from the past few years. In the 2025 session, the 50 states and D.C. introduced over 1,000 AI bills altogether.

The bills during this and recent sessions cover an extraordinarily wide range of targets and approaches. Some bills target AI developers such as Anthropic and OpenAI. Some target deployers of AI such as social media companies or businesses that use AI internally. Others target other parties such as data brokers. Many bills are sector-specific: AI in healthcare, AI in housing, AI in employment, AI in insurance, and AI in elections. And many bills are issue-specific: for example, lawmakers in the 2026 session have introduced 188 bills in 38 states on AI deepfakes and 22 bills in 22 states covering AI chatbots.

A few examples illustrate the range of the 726 bills pending in statehouses and awaiting governor signatures: An Illinois bill would require AI developers to report safety incidents and publicly publish their protocol on risk management, transparency, and cybersecurity (2026 IL SB3312). A Hawaii bill would require AI deployers to run risk management programs for algorithmic discrimination and cybersecurity, including pre-market and ongoing testing, and recordkeeping (2026 HI SB2967). A Minnesota bill would prohibit, “surveillance-based price discrimination,” or the use of AI in using certain consumer data to set prices (2026 MN HF 3764). A New Jersey bill would require companies to conduct AI safety tests and report results to the state (2026 NJ S 1802). A New York bill would hold companies liable for harm caused by AI chatbots offering medical, legal, and other types of regulated speech (2025 NY S7263). 

So far, 13 states this session have enacted or adopted 14 pieces of legislation. A few examples illustrate the range of what lawmakers are passing: Indiana placed restrictions on when healthcare insurance providers can use AI (2026 IN H 1271). New York state and local government may not use AI to reduce staffing, or as the language reads, from using AI in a way that would displace governments jobs (2025 NY S 8831). South Carolina placed restrictions on how data can be collected from minors and implicated AI in the law (2025 SC H 3431). In Vermont, AI videos of political candidates must now be labeled as such (2025 VT S 23).

In a recent Perspectives from FSF Scholars, my colleague, Joe Kennedy, suggests the need for a streamlined AI regulatory framework that incentivizes the build-out of a robust supporting infrastructure and that encourages competition and innovation. What the nation really needs is an overarching federal framework that avoids ex ante heavy-handed regulation and that supplants the growing patchwork of state laws.

Without such a framework, companies must navigate a growing and inconsistent patchwork of state regulations, each with varied definitions, thresholds, compliance timelines, and enforcement mechanisms. States may still decide to pass legislation on AI as it pertains to their specific state criminal codes, public education requirements, state government use of AI, or other state matters. But at the current rate, an AI developer, deployer, or other AI party could theoretically face 51 different pieces of legislation regulating the same activity. And the burden of complying with this patchwork falls even harder on startups and emerging competitors trying to offer better alternatives for consumers. AI has potential to improve countless dimensions of everyday life. The emerging regulate-first patchwork of state laws is not the path to realizing that potential.

The Broadband Providers Growing Role in AI

AI services and platforms and providers of high-speed broadband services are increasingly engaged in a symbiotic relationship. So I argue in the Free State Foundation’s latest Perspectives from FSF Scholars, AI promises significant improvements to broadband providers. But AI itself depends on the networks that collect, analyze, and transmit massive amounts of data. I point out that: “[i]f AI cannot obtain the vast amounts of power and water it needs, if it cannot connect with users to gather data and deliver value, or if access to this infrastructure is compromised, AI collapses.” 

This trend has caught the attention of industry leaders. NCTA President and CEO Cory Gardner recently gave a keynote speech to the State of the Net Conference. He stated that: “[T]his critical infrastructure isn’t just a byproduct of AI. It is the backbone, the workhorse, the foundational element that made and makes AI possible.” This is the result of sustained private investment. According to Gardner, NCTA members have invested more than $355 billion in broadband infrastructure over the last 20 years, including $26 billion last year.

On March 11 President and CEO of CTIA Agit Pai wrote on BroadbandBreakfast that “breakthroughs in artificial intelligence won’t matter much if the networks connecting devices and infrastructure cannot support them. AI without a strong wireless network is like a new car without a road.” Pai pointed to two pillars for success: harmonized spectrum policy and investment-friendly regulatory frameworks.

The dependence on reliable, secure, and high-capacity communications also applies to networks for power, water, and transportation. Each increasingly requires powerful broadband networks to provide the large amounts of data and computing ability required by AI. In order to attract the massive amounts of private investment required, regulators will have to craft sensible regulations that minimize uncertainty and delay. Luckily, the providers of broadband networks are increasingly aware of the opportunity. Hopefully, policymakers are too.

TAKE IT DOWN Act Passed by Congress, Heads to President's Desk

On April 29, the U.S. House of Representatives passed, by a 409-2 vote, the Tools to Address Known Exploitation by Immobilizing Technological Deepfakes on Websites and Networks Act or the "TAKE IT DOWN Act" (S. 146). The bill, which passed by unanimous voice vote in the Senate on February 13, now goes to President Donald Trump's desk for signature. 

As described in a January 16 press release by the Senate and House bills' sponsors, the TAKE IT DOWN Act “makes it unlawful for a person to knowingly publish [non-consensual intimate imagery (NCII)] on social media and other online platforms. NCII is defined to include realistic, computer-generated pornographic images and videos ["deep forgeries"] that depict identifiable, real people." The bill has separate provisions and corresponding criminal penalties applicable to minors and adults, and it specifies that a victim consenting to the creation of an authentic image does not mean that the victim has consented to its publication.

 

Additionally, the TAKE IT DOWN Act includes a notice-and-takedown provision that requires social media and other public websites or internet services to establish procedures for the removal of NCII in response to a valid request from a victim, within 48 hours. Under the bill, websites also are required to make reasonable efforts to remove copies of the unauthorized images. Websites that make good faith efforts to remove NCII or disable access to it receive immunity from legal claims relating to such removal or disabled access. However, a website’s failure to comply with the notice-and-takedown requirements constitutes an unfair or deceptive act or practice under the Federal Trade Commission Act. Under the TAKE IT DOW ACT, the FTC has authority to enforce the notice-and-takedown requirements and impose penalties for non-compliance. 

 

The TAKE IT DOWN Act appears to be a commonsense measure, carefully written, and reasonably necessary to address a serious problem that is nationwide in scope. President Donald Trump is expected to sign the bill into law. Credit and congratulations are due to the bill's supporters and its sponsors.

 

The TAKE IT DOWN Act (S.146) is sponsored by Senators Ted Cruz and Amy Klobuchar. Reps. Maria Elvira Salazar and Madeleine Dean are sponsors of the House companion bill (H.R.633). Senator Cruz, who is Chairman of the Senate Commerce, Science, and Transportation Committee, talked about the TAKE IT DOWN Act during his keynote address at the Free State Foundation's Seventeenth Annual Policy Conference in Washington D.C. on March 25, 2025: 

NO FAKES Act to Combat "Deepfakes" is Reintroduced in Congress

On April 11, the "Nurture Originals, Foster Art, and Keep Entertainment Safe Act of 2025" or "NO FAKES Act" was re-introduced in the U.S. House of Representatives (H.R. 2794) and Senate (S. 1367). The House bill is sponsored by Rep. Maria Elvira Salazar and the Senate bill is sponsored by Sen. Christopher Coons. The NO FAKES Act would bolster individuals' intellectual property rights in their likenesses and voices by recognizing a private right of action against unauthorized and harmful "deepfakes." The bill has bipartisan backing as well as the endorsement of a cross-section of the creative and tech industries. The NO FAKES Act is strong on the merits and the 119th Congress should give it due consideration. 

 

Although generative AI technologies offer potential benefits, they also may be abused. Public displays and dissemination of "deepfake" songs misappropriate the value of recording artists’ voices, damaging the artists economically. Also, generative artificial intelligence (AI) tools and services on the Internet allow users to create "deepfake" explicit pictures and videos of individuals.

 

The NO FAKES Act would address those "deepfake" dangers in a targeted way by establishing a national uniform baseline of legal protection for an individual’s likeness and voice from unauthorized digital replicas. If passed by the 119th Congress and signed into law by President Donald Trump, the Act would make civilly liable anyone who knowingly produces a digital replica without the consent of the rights owner. It also would make civilly liable anyone who knowingly publishes, reproduces, displays, distributes, transmits, or makes the digital replica available to the public without the rights owner's consent. Persons harmed under the Act would have a right to seek statutory or actual damages, recovery of costs and attorneys’ fees, and injunctive relief. 

 

Recognizing the potential benefits of authorized digital replicas, the NO FAKES Act provides that individuals would have the right to license their personas for digital replication by third parties. Additionally, the Act is carefully written to address abuses and it includes safeguards for First Amendment-protected free speech and expression using generative AI tech. It bears emphasis that the NO FAKES Act is about private law – personal rights and intellectual property rights; it is not a federal criminal law bill.

 

A more detailed review of the same bill, previously introduced in the 118th Congress, is provided in my August 2024 Perspectives from FSF Scholars, "The 'NO FAKES Act' Would Protect Americans' Rights Against Harmful Digital Replicas."

Will AI Help or Hinder Federal Privacy Legislative Efforts?

Efforts to pass a federal data privacy law have dragged on for many years. During that time, unrelenting technological advancement simultaneously has produced new innovations that amplify calls for clear rules and complicated congressional conversations that might lead to such legislation. Artificial Intelligence (AI) is the latest such instigator/troublemaker.

Generative AI offerings – such as OpenAI's ChatGPT, Google's Gemini, and Meta AI – depend upon Large Language Models (LLMs) trained on massive amounts of data. The more data used to train the LLM, the better the results. Consequently, generative AI raises substantial questions relating to privacy. (By way of example, the image below was created with OpenAI's DALL-E using the prompt "create an image of generative AI and data privacy.")

In her Opening Statement regarding a recent Senate Commerce, Science and Transportation Committee hearing titled "The Need to Protect Americans' Privacy and the AI Accelerant," Chair Maria Cantwell (D-WA) wrote that "[w]e are being surveilled … tracked online in the real world, through connected devices. And now, when you add AI, it is like putting fuel on a campfire in the middle of a windstorm." AI, she argued, "increases the need for passing legislation soon."

This heightened concern, however, to date has not generated legislative progress on data privacy. The American Privacy Rights Act of 2024, about which I wrote in "Congressional Leaders Return Privacy to the Front Burner," an April 2024 Perspectives from FSF Scholars, has yet to advance beyond a discussion draft. It was scheduled for markup by the House Energy and Commerce Committee on June 27, 2024, but that markup was cancelled at the last minute, a development I described in a post to the Free State Foundation's blog.

Prompting an unsettling sense of déjà vu, already one state has taken stalled congressional matters into its own hands. On May 17, 2024, Colorado Governor Jared Polis signed into law Senate Bill 24-205, "Concerning Consumer Protections in Interactions with Artificial Intelligence Systems."

Broadly speaking, Senate Bill 24-205, which goes into effect on February 1, 2026, requires that developers of "high-risk" AI systems "use reasonable care to protect consumers from any known or reasonably foreseeable risks of algorithmic discrimination."

We shall see if other states follow Colorado's lead – and, if so, whether another unwanted privacy-related "patchwork" emerges.

Copyright and AI

Both real-world advances and outright speculations about artificial intelligence (AI) technologies have prompted scholars, policymakers, and others to ponder the implications of AI for copyright law and policy. The Copyright Office is hosting a symposium on "Copyright in the Age of Artificial Intelligence" on February 5 at the Library of Congress. And in 2019, the U.S. Patent and Trademark Office requested public comments on the impact of AI on copyrights and other forms of intellectual property (IP). Among the comments filed in response to the USPTO's request, the Motion Picture Association (MPA) offered a common sense take on the durability of basic copyright principles and the importance of maintaining clear rules regarding ownership and liability for infringement.

We may have more to say on AI and copyright the future. However, Free State Foundation President Randolph May and I have made the case – in Perspectives from FSF Scholars papers published in September 2019 and here in January 2020 – that copyright infringement is a strict liability tort and that an online platform provider's use of an automatic process in causing an infringement does not shield such providers from liability.