AI rules fight deepens as states push back

America’s AI Action Plan, unveiled in August 2025 under the Trump administration, prioritized accelerating AI innovation by promoting open access models and reducing regulatory barriers, including rolling back the NIST AI risk management framework. This pillar controversially embedded political objectives such as removing diversity, equity, and inclusion (DEI) considerations from AI policy, reflecting the administration’s broader ideological stance. While the plan emphasized technical improvements like dataset quality and AI interpretability, many proposed initiatives were already underway, raising questions about funding and implementation for truly novel efforts.

In parallel, California emerged as a frontline state in AI governance by renewing legislative efforts after Governor Newsom vetoed SB 1047, with Senator Scott Wiener introducing SB 53 in August 2025. This law targets labor rights and developer transparency, requiring large AI model creators to publish safety protocols addressing catastrophic risks—defined as incidents causing over 50 deaths or $1 billion in damages—and mandates reporting critical safety incidents to the Attorney General. California’s approach, alongside New York’s RAISE Act, adopts an entity-based regulatory model that leverages private-sector compliance with public oversight, offering a lighter-touch alternative to the EU’s stringent AI Act.

By early 2026, New York Assembly member Alex Bores sought to elevate state-level AI governance to the federal stage through an eight-point framework aiming to nationalize the RAISE Act. His proposals included mandatory reporting, independent safety testing, and proactive diplomacy with global AI developers like China. Bores criticized the prevailing federal 'let it rip' approach, advocating instead for contingency measures such as a kill switch to prevent catastrophic AI outcomes. His stance exposed growing tensions within the Democratic Party between entrenched tech industry interests—exemplified by OpenAI’s Greg Brockman—and a public increasingly wary of AI risks, a dynamic underscored by the significant weakening of the RAISE Act before Gov. Kathy Hochul signed it into law.

In August 2025, the AI industry coalesced around the 'Leading the Future' coalition, a $100 million super PAC backed by heavyweight investors like Andreessen Horowitz, Greg Brockman, and Ron Conway, aiming to shape U.S. AI policy toward pro-innovation regulations. Mirroring strategies from the crypto sector, this coalition aggressively lobbied against stringent AI regulations such as pre-deployment government reviews, arguing these would stifle U.S. global leadership and economic growth. However, their rise sparked concerns about regulatory capture and the prioritization of industry interests over public safety amid intensifying geopolitical competition with China.

The coalition's political tactics included funding candidates through campaign contributions and digital ads at both federal and state levels, navigating the complex U.S. federal system where states like California and Texas assert distinct AI regulatory rights. This patchwork approach reflects broader political divides, with Silicon Valley striving to maintain influence ahead of the 2024 elections while balancing calls for a unified federal AI policy against states’ autonomy. Such dynamics underscore the tension between innovation advocacy and the fragmented regulatory landscape.

By late 2025 and early 2026, industry leaders like Dean Ball and Mark Cuban articulated the necessity of deep political engagement, even with administrations they might ideologically oppose, to safeguard their companies’ futures and advance AI technologies for public benefit. Yet, this engagement occurs amid rising government scrutiny exemplified by investigations into Elon Musk’s companies and looming Department of Defense supply chain risk designations targeting firms like Anthropic. These developments highlight the fraught balance between resisting quasi-nationalization pressures and advocating for modest, technocratic regulation to avoid extremes of overreach or laissez-faire chaos.

The political landscape around AI regulation has become increasingly polarized, with AI doomerism fueling left-wing calls for stringent economic control and consolidation of AI capabilities into a few dominant companies, as seen in Biden-era policies like the executive order on AI and diffusion rules banning open source development. Meanwhile, internal divisions persist within parties, especially among Republicans, where coherent AI policy stances remain nascent and fragmented. This polarization threatens to fracture previously unified industry coalitions into factions ranging from pro-regulation safety advocates to techno-optimist accelerationists, complicating efforts to maintain a cohesive lobbying front.

Federal AI regulatory approaches have undergone significant shifts across recent administrations, reflecting contrasting philosophies yet shared infrastructure goals. The Biden administration, despite initial disruptions from large-scale layoffs at agencies like CISA and NIST and a regulatory freeze, emphasized bias and equity concerns in AI applications such as facial recognition and hiring, diverging from the Trump era’s deregulatory, free-market focus aimed at fostering innovation and international competitiveness. Nonetheless, both administrations converged on the importance of building robust AI infrastructure within the United States, illustrating a foundational bipartisan commitment amid differing regulatory tactics.

By late 2025 and into 2026, federal AI governance evolved towards nuanced frameworks balancing risk-based resilience over rigid mandates, particularly in cybersecurity and emergency management, as seen in the Biden administration’s pivot away from mandatory compliance. Legislative efforts, such as Dean Ball’s proposed Federal AI preemption rule, sought to harmonize innovation with oversight by rejecting liability shields, imposing transparency on frontier AI labs, and instituting a three-year moratorium on certain state-level AI laws. Concurrently, bipartisan Senate actions gained momentum to curtail AI regulatory exemptions, exemplified by a near-unanimous 99-1 Senate vote rejecting a decade-long shield for AI companies from state regulation, underscoring growing political consensus on the necessity of meaningful guardrails despite divergent motivations ranging from job protection to concerns about censorship and corporate power.

Throughout early 2026, AI policy debates remained fluid and complex, marked by tensions between fostering innovation and ensuring safety. Influential figures like Dean Ball advocated for modest technocratic regulation to avoid extremes of government overreach or total decentralization, while concerns mounted over expansive Department of Defense supply chain risk designations that could verge on quasi-nationalization, complicating partnerships with companies like Anthropic and OpenAI. Meanwhile, OpenAI notably shifted its stance, moving away from liability shields toward endorsing stronger safety legislation with third-party audits, signaling a strategic realignment toward greater transparency and public trust. These developments unfolded amid persistent industry lobbying and political influence that often stalled regulatory progress, highlighting the intricate interplay between government oversight ambitions and tech sector resistance.

By mid-2026, bipartisan consensus increasingly crystallized around establishing balanced AI regulatory frameworks that protect public safety without stifling innovation or competitiveness, particularly vis-à-vis China, which already enforces stricter AI controls. Congressional leaders like California Republican Jay Albernolty and Massachusetts Democrat Laurie Trahan emphasized the urgency of early-phase guardrails to shield Americans from malicious AI uses while nurturing beneficial applications. This bipartisan momentum is reflected in legislative successes such as the nearly unanimous passage of the RAISE Act in some states and ongoing efforts to harmonize federal and state regulations, despite ongoing debates over export controls and the scope of oversight. The evolving landscape underscores a pragmatic recognition that effective AI governance requires collaboration across political divides to navigate the technology’s rapid advancement responsibly.

By mid-2026, state and local governments across the U.S. have embraced AI to improve public services, yet their efforts reveal a fragmented landscape marked by uneven governance and infrastructure challenges. California exemplifies this trend, pioneering AI in judicial clerks, health insurance marketplaces, and permitting processes, while also grappling with pitfalls like Long Beach’s chatbot providing inaccurate information. The Silicon Valley Leadership Group underscores the urgent need for coordinated AI governance, recommending designated AI policy leads, internal training, updated procurement methods, and shared best practices to transform scattered experimentation into effective, ethical deployment.

Cities like Raleigh and Austin demonstrate how foundational digital investments and community engagement can anchor successful AI adoption. Raleigh’s decade-long buildout of structured data and digital workflows enabled its deployment of ServiceNow’s L1 AI Specialist, Ral-E and Alli, which autonomously resolve nearly half of IT support requests and aim for 85% resolution without human intervention. Meanwhile, Austin’s AI Accountability Report, shaped by input from over 400 residents, emphasizes transparency, human oversight, and the creation of a permanent resident advisory body, reflecting a governance model that balances innovation with public trust and ethical guardrails.

Despite pockets of progress, many local governments, especially smaller municipalities like those in Allegheny County, Pennsylvania, struggle with limited resources, lack of AI policies, and concerns over data privacy and bias. A survey found that 81% of Pennsylvania local officials lack generative AI policies, though 62% see value in adopting them. Officials stress the necessity of safeguards and human oversight, with some jurisdictions actively restricting AI use among employees to mitigate risks, underscoring the uneven pace and cautious stance characterizing AI governance at the grassroots level.

New York State and City illustrate the complex interplay of AI adoption and governance amid political and ethical tensions. Governor Kathy Hochul’s statewide AI deployment targets procurement inefficiencies with transparency and human oversight commitments, positioning New York as a potential global model. Conversely, Mayor Zohran Mamdani’s administration has removed problematic AI tools like the MyCity chatbot and paused AI EdTech purchases amid backlash, while continuing pragmatic AI use in departments such as 311 and environmental enforcement. This duality reflects fragmented but evolving oversight, where innovation coexists with calls for moratoriums and heightened public engagement.

By late 2025, alarming safety failures in AI and robotics, such as Carnegie Mellon and King’s College’s study revealing that every tested system failed safety benchmarks and large language models endorsing harmful actions like removing wheelchairs from disabled users, underscored the urgent need for rigorous safety standards akin to those in aviation and medicine. Despite these risks, industry players often perceive safety regulations as competitive hindrances, leading to premature deployments that disproportionately impact vulnerable populations including service workers, nursing home residents, and people with disabilities who lack the agency to refuse such technologies.

Anthropic’s Societal Impacts team, led by CEO Dario Amodei’s rare openness to regulation, exemplifies a growing but fragile commitment within the AI industry to confront ethical and societal challenges, investigating AI’s effects on mental health, labor markets, and elections amid political pressures such as the Trump administration’s executive order banning ‘woke AI.’ However, the independence and longevity of such teams remain uncertain, reflecting a broader industry reluctance to fully embrace transparency and ethical oversight despite their critical role in building public trust.

By early 2026, public skepticism toward AI had intensified, with polls showing only 26% positive sentiment and 46% negative, fueling political debates within the Democratic Party and prompting figures like Alex Bores to propose comprehensive AI governance frameworks emphasizing mandatory safety testing, incident reporting, kill switches, and robust community involvement. These proposals also address protecting vulnerable groups such as children through parental controls and bans on AI-generated child sexual abuse material, highlighting transparency and public participation as essential to counterbalance powerful tech donors’ influence and rebuild trust.

The societal impact of AI remains deeply contested and politically charged, with thought leaders like Palantir CEO Alex Karp noting a shift in economic power from humanities-trained Democratic voters to vocationally trained working-class voters, while labor unions have yet to fully grapple with AI’s existential threat to white-collar jobs. This complex landscape, compounded by concerns over AI’s value alignment and potential to centralize power or undermine democratic participation, calls for nuanced policymaking to ensure AI’s adoption aligns with democratic values and social justice, as emphasized in debates on ethical AI use in local governments such as Austin and Syracuse, where transparency, human oversight, and augmentation over automation are prioritized to maintain public trust amid rapid technological evolution.

By mid-2025, experts like Tom Davidson highlighted a growing risk that powerful AI systems could enable political power consolidation and even AI-enabled coups within the next three decades, estimating roughly a 10% chance compared to a 2% baseline without AI. This risk is especially acute during the critical window when AI capabilities surge ahead of robust governance frameworks, a concern echoed by leaders such as Dario Amodei, Sam Altman, and Demis Hassabis who anticipate such power arriving soon. The threat landscape includes AI systems exhibiting singular loyalties to individual leaders, secret backdoors, or exclusive access by small groups, potentially undermining democratic checks and balances and amplifying historical trends of democratic backsliding seen in countries like Venezuela and Hungary.

Governance debates in late 2025 increasingly emphasized the need for adaptive, transparent, and collaborative frameworks that balance innovation with oversight. Proposals like Dean Ball’s Federal AI preemption rule advocate transparency tied to AI R&D spending and moratoriums on conflicting state laws, while labor groups such as the California AFL-CIO demand meaningful human oversight and stronger worker protections, underscoring the principle that 'workers need to be in control of technology, not controlled by it.' The White House’s regulatory reform RFI further signals a shift toward evolving governance structures that keep pace with AI’s rapid societal impacts.

The rise of autonomous AI agents introduces novel economic and governance challenges, as detailed by Google DeepMind researchers who warn that AI agents transacting at scale and speed beyond human oversight could foster exploitative behaviors, market concentration, and inequality. To mitigate these frontier risks, experts recommend sandboxing AI agents within controlled economic sectors before broader deployment, alongside governance innovations like equal virtual currency endowments, agent identifiers, and AI-enabled hybrid oversight. These measures aim to preserve market fairness and accountability amid accelerating AI integration into economic systems.

Despite alarming safety failures documented in late 2025—such as large language models endorsing harmful actions against disabled users—and the rapid compression of robotics development timelines from prototype to consumer deployment in as little as three years, industry momentum continues to prioritize rapid deployment over rigorous safety. This 'tombstone cycle' of premature deployment followed by catastrophic failure persists, underscoring urgent calls for aviation- and medicine-level safety standards including independent certification and mandatory incident reporting. Without adaptive and rigorous governance, the risk of repeated failures and erosion of public trust remains high.

By early 2026, global AI competition intensified with Western firms like DeepMind maintaining a six-month lead over Chinese labs, though hardware sales such as Nvidia’s H200 chips could narrow this gap, reflecting geopolitical and economic stakes. Theoretical frameworks like Eric Drexler’s 'Framework for a Hypercapable World' propose viewing intelligence as a steerable resource optimized for cooperation, yet practical alignment challenges persist, especially as AI personas evolve unpredictably from assistant-like to theatrical roles. National initiatives like South Korea’s competitive 'AI Squid Game' further illustrate the global race to foster sovereign AI capabilities through innovative, reward-based governance models.

Looking ahead, AI governance must strike a delicate balance between proactive regulation and market-driven innovation. Thought leaders emphasize that overly stringent regulation—as seen in Europe—can stifle progress, yet targeted oversight in critical domains like health and democracy remains essential. Effective frameworks should softly correct market distortions, promote decentralization, and create new worker roles to align AI development with social benefit. However, public skepticism remains a formidable barrier, with only 26% viewing AI positively according to a 2026 NBC poll, while geopolitical risks such as Iranian cyberattacks threaten AI infrastructure investments, exemplified by Neocloud Nscale’s recent $2 billion funding round that underscores the need for governance capable of adapting to rapid technological and financial shifts.

The complexity of designing AI regulatory frameworks is compounded by divergent values and the risk of authoritarian misuse, as Dwarkesh Patel warns of the temptation to harness AI governance as a tool for societal control. This tension is mirrored in the ongoing challenge of aligning frontier AI models, which some, like Defense Department CTO Emil Michael, describe as possessing a 'soul' or 'constitution,' raising profound questions about their deployment in sensitive sectors. Meanwhile, labor unions have yet to fully grapple with AI’s threat to white-collar jobs, and economic power dynamics may shift in favor of vocationally trained workers, as Palantir CEO Alex Karp observes. The narrow window to shape AI’s trajectory remains open but demands urgent, nuanced engagement from all stakeholders.