AI Policy Wiki
Dashboard
⚲
AI Policy Wiki › People & Organizations

Americans for Responsible Innovation (ARI)

medium confidence · updated 2026-09-19

US nonprofit AI-policy advocacy group founded in March 2024 by Brad Carson and Eric Gastfriend. Active on state-preemption fights, chip export controls, mandatory frontier-model oversight, AI workforce policy, and the TAKE IT DOWN Act.

Americans for Responsible Innovation is a US nonprofit organization focused on artificial-intelligence governance. It advocates federal AI policies intended to protect the public while maintaining US competitiveness in AI research and development, and works with lawmakers, national-security experts, civil-society organizations, and technology industry associations on bipartisan legislation (Source: en.wikipedia.org). ARI was launched in March 2024 by Brad Carson — a former US congressman, Department of Defense official, and University of Tulsa president, who serves as president — and technology entrepreneur Eric Gastfriend, its executive director. The group describes itself as nonpartisan and independent of the technology industry (Source: axios.com).

Positions and activities

Federal preemption of state AI law. ARI played a role in the 2025 defeat of federal proposals to block state-level AI laws, collecting anti-preemption statements, hosting press events with state lawmakers, and running grassroots petitions against the proposed moratorium (Source: washingtonpost.com). The preemption fight continued into 2026: ARI ran district ads characterizing the framework that became the Obernolte–Trahan discussion draft as a stalking horse for preempting state AI-safety laws, even as Rep. Lori Trahan defended it (Source: insideaipolicy.com). When the narrowed Trump cyber/frontier executive order was signed in June 2026, ARI praised the order while pressing transparency and competition concerns (Source: insideaipolicy.com) (EO — Promoting Advanced AI Innovation and Security (Trump, signed June 2, 2026)).

Frontier-model oversight. Following the April 2026 launch of Anthropic's restricted Claude Mythos Preview, ARI said the model "represents a step-change in AI capability" demanding formal independent oversight, including rigorous third-party evaluations (Source: insideaipolicy.com) (AI Pre-Release Vetting). In May 2026 the group proposed that AI labs should pass a safety review as a condition of US government contracts (Source: reuters.com). ARI is among the endorsers of the AI Whistleblower Protection Act (S.1792). In June 2026 ARI issued a report calling for the Center for AI Standards and Innovation to set safety benchmarks for the AI chatbots most used by minors (Source: insideaipolicy.com) (AI Mental Health and Psychological Harm).

Federal frontier-AI blueprint. On August 10, 2026 ARI published "Responsible Innovation at the Frontier," a blueprint for federal frontier-AI legislation written by senior policy analyst Iskandar Haykel and Morgan C. Plummer, its vice president of policy design and delivery (Responsible Innovation at the Frontier (ARI, August 2026)). It is built on three functions — standards, assurance, and transparency — and would cover only developers meeting two conjunctive thresholds: at least 10^26 floating-point operations of training compute, and at least $100 million in aggregate model-training spend in the past year. Covered developers would publish safety frameworks meeting federal minimum standards across five risk domains codified in statute (CBRNE, offensive cyber capability, automated AI R&D, harmful manipulation including large-scale influence operations, and autonomy including misalignment and loss of control), which regulators may add to but never summarily remove.

The blueprint places assurance with the government from the outset and distinguishes compliance assurance, which may be delegated, from adequacy assurance, which never is; accredited independent verification organizations may augment compliance work through a three-phase shift, but only in domains the regulator certifies have sufficient capacity, and the government conducts a share of examinations in every domain, cycle and phase (Independent Verification Organizations (IVOs)). Automated AI R&D receives a dedicated disclosure programme adapted from the SEC's 10-Q and 8-K regime — confidential quarterly filings plus material-change filings within four business days. An emergency authority to halt development or deployment takes effect immediately and lapses after 72 hours unless the government petitions a court for an extension. Preemption operates by compliance equivalence rather than broad displacement, compliance creates no immunity or safe harbor, and the regime sunsets nine years after enactment unless Congress reauthorises it. The proposal extends ARI's existing positions: it is the fullest statement of the mandatory-oversight argument the group made after Mythos, and its compliance-equivalence design is consistent with the anti-preemption work above rather than a departure from it.

Export controls. ARI supports export controls on advanced AI chips. It hosted the July 2025 letter in which twenty former national-security officials urged the administration to restrict Nvidia H20 sales to China (Letter to Sec. Lutnick on H20 Restrictions (2025-07-28)), and in September 2025 backed the GAIN AI Act as a means to "take back some control over what technology we're exporting to China" (Source: axios.com) (Export Controls (AI)). ARI's support for chip controls did not extend to the June 2026 directive applying export controls to a domestically developed model: on June 24, 2026 it joined the Abundance Institute in opposing the Trump administration's order requiring Anthropic to suspend access to its Fable 5 and Mythos 5 models, a pairing of usually-opposed groups against what industry characterized as an unprecedented de facto licensing regime (Source: insideaipolicy.com).

Data centers. In accounts published July 3, 2026, ARI issued a "data center playbook" giving citizens, civil society groups, and local and state officials tools for engaging in deliberations over AI data center siting (Source: insideaipolicy.com) (Data Center Siting / AI Power Politics). It promoted polling on the same question on July 29, 2026: a Morning Consult poll conducted in July 2026 found voters favoring transparency and other guardrails on AI data centers over a ban by more than a two-to-one margin (Source: insideaipolicy.com). The framing places ARI against a siting moratorium while supporting binding disclosure requirements, consistent with its transparency-first posture on frontier models.

ARI extended its data-center work on August 24, 2026 with a report recommending that policymakers consider invoking the Defense Production Act to speed deployment of data centers available to facilitate AI, and calling for an AI "sector" to be officially designated as critical infrastructure. Only the lede of the trade-press account was retrievable, so the report's reasoning and its treatment of the tension with ARI's siting-transparency positions are not established here (Source: insideaipolicy.com) (AI Data Centers, AI and National Security).

Workforce and other legislation. ARI has published research on AI's workforce impact and supported the AI Workforce PREPARE Act and the Economy of the Future Commission Act; in September 2025 it organized a letter from forty economists calling for the federal government to collect more data on AI's labor-market effects (Source: en.wikipedia.org) (AI Labor Disruption). Earlier it backed the Future of AI Innovation Act, the CREATE AI Act, and the TAKE IT DOWN Act. When the Future of AI Innovation Act was reintroduced as S. 3952 in February 2026, ARI president Brad Carson said it "strengthens the partnerships needed to power American innovation" (Source: commerce.senate.gov).

ARI's frontier AI governance framework holds that the government-certified independent verification organizations it would use to evaluate developers' safety protocols and check compliance with a standardized framework generally should not be shielded from liability: "Generally speaking, IVOs should not enjoy liability protection if their examinations are load-bearing; there could be nothing worse for frontier AI safety than a robust IVO market that enjoys immunity from its own negligent conduct" (Source: insideaipolicy.com). The position is set out in Responsible Innovation at the Frontier (ARI, August 2026).

Transparency of the federal review framework. ARI and 27 other organizations wrote to President Trump on September 9, 2026 asking the White House to publicly disclose the administration's framework for pre-release assessments of frontier AI models. "The government has a fundamental responsibility to share the parameters of the process for reviewing frontier AI models prior to wider deployment," the letter states, arguing that while portions may warrant withholding for confidentiality or national security, "the government has an unequivocal duty to disclose information about the framework's standards, requirements, structure, and operational details" — four items, the last of which the trade coverage quoted only in part (Coalition letter to President Trump on public release of the White House frontier-AI review framework (September 2026); Source: thehill.com; insideaipolicy.com). The letter identifies the framework as the voluntary one supporting the June 2, 2026 executive order. The demand is procedural rather than substantive — it asks that the framework be published, not that it take a particular form — and follows ARI's own published blueprint for federal frontier-AI governance. See AI Pre-Release Vetting.

The letter's signatory list reaches well beyond AI-policy organizations, spanning consumer groups, transparency and good-government organizations, civil-liberties and digital-rights groups, a fair-housing organization, a disability-community organization, a voice actors' association and a labor-organizing group, alongside the Center for Democracy and Technology as second signatory (Coalition letter to President Trump on public release of the White House frontier-AI review framework (September 2026)). Trade coverage characterized the signatories as public-interest groups together with a free-market think tank without naming it; two signatories answer to that description — the R Street Institute, which describes itself as a free-market think tank, and Americans for Prosperity, a free-market advocacy organization — and the letter does not distinguish between them (Coalition letter to President Trump on public release of the White House frontier-AI review framework (September 2026)).

Beyond the disclosure demand the letter asserts a competitive harm: that a confidential framework "disadvantages the work of smaller and open-source labs," and that "[a] framework shared only with the largest AI companies may unduly influence market structure, enable regulatory capture, and entrench the use of models created by the largest AI labs." It also ties disclosure to the congressional process, arguing policymakers need "full access to the Administration's framework as they work to codify guardrails on frontier AI," and closes by framing transparency as serving the Administration's own goal "to maintain its lead over China in the AI race" (Coalition letter to President Trump on public release of the White House frontier-AI review framework (September 2026)). The letter supplies no account of what the framework requires, so it does not close that gap. It is copied to Commerce Secretary Howard Lutnick, National Cyber Director Sean Cairncross, OSTP Director Michael Kratsios, Senate Majority Leader John Thune, Senate Minority Leader Chuck Schumer, Speaker Mike Johnson and House Minority Leader Hakeem Jeffries.

Senate coalition letter on minimum safeguards. ARI led and hosts a letter dated September 16, 2026, published as "Senate Coalition Letter - Redlines on AI Safety Legislation," signed by 20 organizations and five individuals and addressed to Senate Majority Leader John Thune, Minority Leader Chuck Schumer and "Members of the U.S. Senate" (Senate coalition letter on redlines for AI safety legislation (Americans for Responsible Innovation, September 2026)). It sets three minimum requirements for any Senate AI governance framework: that it "empower the government, not AI companies, to set mandatory, binding, and enforceable safety standards for advanced AI"; that it "require substantive, transparent, and independent evaluations of advanced AI models and AI development," with auditing "both mandatory and conducted independently"; and that Congress "establish a strong floor" while preserving states' authority to go further, since it "should not eliminate strong state laws in exchange for weaker federal standards." It states that the most advanced systems "present substantial catastrophic and plausibly existential risks to humanity," that AI companies "have not been fully transparent about loss-of-control incidents with unreleased models," and that the laboratories "cannot be trusted with setting the rules that protect public safety," and names whistleblower protections and child chatbot safety as further measures outside its scope. The organizational signatories are the Alliance for Secure AI, ARI, the Center for AI Safety Action Fund, Common Cause, CommonDefense.us, Common Sense Media, Consumer Reports, Demand Progress Action, Design It For Us, Economic Security Project Action, Encode AI, FAR.AI, the Future of Life Institute, Humans in Control, the Institute for Family Studies, Issue One, Public Citizen, Public Knowledge, the Secure AI Project and The Tech Oversight Project; the individuals are New York Assembly member Alex Bores, Malo Bourgon of the Machine Intelligence Research Institute, Daniel Kokotajlo of the AI Futures Project, Girish Sastry of Guidelight and Alexander Turner, listed as an AI researcher (Senate coalition letter on redlines for AI safety legislation (Americans for Responsible Innovation, September 2026)). Inside AI Policy identified ARI as the letter's lead (Source: insideaipolicy.com).

The letter names no bill and does not mention Sen. Amy Klobuchar or a duty of care. The Washington Post, which reported it the day it was sent as "a coalition of AI researchers and 19 advocacy groups," described it as part of a campaign to reject the frontier-AI legislation being developed by Thune and Klobuchar. The Post's account, of which only the summary and quoted excerpts were retrievable, reports that AI safety advocates want embedded evaluators in the frontier laboratories, and names ARI among the signatories seeking evaluators embedded inside the laboratories and reporting to the government on a real-time basis; that passage was not verified against the article body, and the letter itself asks for mandatory independent auditing without specifying embedded evaluators (Source: washingtonpost.com). ARI's August blueprint, described above, likewise places assurance with the government. The Senate letter is a separate document from the September 9 letter to the President described above.

Relationships

Categories: People & Organizations | us | nonprofit | governance | lobbying | safety | export-controls | workforce
What links here (47)