Policy Bytes
The latest policy bytes from Americans for Responsible Innovation.
The Pentagon Needs AI Talent, but Can’t Attract & Keep It with Yesterday’s Bureaucracy
The Pentagon took off sprinting in a race to accelerate the adoption of artificial intelligence in our defense ecosystem over the first five months of this year. The Department of War was clear about ...
Minutes to Seconds to Machines Running the Whole Show
A naval intelligence officer aboard a destroyer in the South China Sea watches her console populate. The fused AI targeting system has correlated satellite imagery, signals intelligence, and pattern-o...
After Mythos
Anthropic's latest AI model, Claude Mythos, has arrived with visions of impending upheaval and future crises in tow. Within days of the new model's launch, senior federal financial officials held a sp...
The New Iron Triangle
The Iron Triangle of Painful Tradeoffs
For decades, defense acquisition has operated under a simple constraint known as the iron triangle. Any program can optimize for at most two of three ...
The recent legal victories against Big Tech are promising, but not enough to protect kids
A series of legal victories against Meta captured headlines as families won redress for harms to minors linked to social media and AI products. These cases tested a consequential legal question: wheth...
Wargaming Needs AI. The Department of War Isn’t Ready.
Scene-Setting
You've been invited to an exclusive wargame, held in the hardest-to-access pentagonal-shaped building in the world. You enter an LED-lit room, wrapped in a floor-to-ceiling map dotte...
We Can’t Manage What We Can’t Measure: Why It’s Critical to Standardize Data Center Energy Reporting
While tech titans race to dominate AI benchmarks, their ambitions run on electricity, and lots of it. Take xAI's Colossus supercomputer in Tennessee. It's drawing 150 MW from the grid (enough to power...
Midway to Milliseconds: Wargaming the Next Technology Revolution
Between the world wars, military establishments confronted a technological revolution. Sonar. Radar. Aviation. Submarines. Wireless communications. Each represented a fundamental reordering of how cou...
AI Safety Research Highlights of 2025
2025 has shaped up to be a watershed year for AI safety and security. Frontier models showed improved potential to facilitate CBRN threats, lending greater salience to these catastrophic risks. Leadin...
Five AI Policy Wins in the NDAA
Congress is laying the groundwork for the deployment of artificial intelligence by mandating its responsible use across America's defense infrastructure. Buried in the thousands of pages of the FY26 N...
Beyond Export Controls: Addressing Key National Security Issues Related to AI
The recent release of the Trump Administration’s AI Action Plan, as well as more than 30 provisions related to AI between both the House and Senate versions of this year’s National Defense Authori...
A Guide to AI Interpretability
Introduction
If a Large Language Model (LLM) used as a healthcare chatbot offers reassurance instead of flagging a serious symptom, can we trace what drove that decision? What if a video generation...
The Weakest Link: Strategic Inputs in U.S.-China AI Competition
While most attention paid to U.S.-China AI competition focuses on advanced compute chips used to scale models, an area where the U.S. currently has the advantage, there are other key inputs to AI deve...
Securing AI Through Defense in Depth: The Challenge of Jailbreaks
On May 22, 2025, internal discussion on Anthropic's new Claude Opus 4 caused the leading AI developer to trigger AI Safety Level 3 (ASL-3)—a capability threshold indicating the models may significan...
Seven Key Features of Trump’s AI Plan
The strengths and risks of the White House’s AI Action Plan
The AI Action Plan released by the White House on Wednesday seeks to secure America’s leadership in AI through infrastructure investm...
State-Level AI Laws May Drive Adoption of AI Tools
State-level AI legislation does not hinder public interest in and adoption of generative AI tools—in fact, it may help drive it. As policymakers weigh the risks and benefits of AI regulation, these ...
Reward Hacking: How AI Exploits the Goals We Give It
What happens when AI systems learn how to cheat?
In April 2025, OpenAI released system cards for their o3 and o4-mini “reasoning” models detailing technical specifications as well as safety and...
Future-Proofing Governance in the Age of AGI
A century after Mr. Smith went to Washington, artificial intelligence is on its way to town. Over the past year alone, we’ve seen the creation of Chief AI Officers at federal agencies, and now Elon ...
AI Security Tax Incentives
The Case For AI Security Tax Incentives & How They Could Work
Americans for Responsible Innovation recently released a proposal to institute a tax incentive for research on AI security and resp...
Explainer: DeepSeek, Distillation, and AI IP Theft
What Distillation and Model Weight Theft Means for the Future of U.S. AI Innovation & Policy
Last week, ARI held a panel with technical and policy experts on AI model piracy – stealing AI IP ...
Policy Recommendations: Full Funding for NIST and AI Incident Reporting
Today, ARI is releasing two new policy memos for the Trump administration and the 119th Congress as they develop their AI policy priorities. Implementing the memos’ primary recommendations – fulfi...
Explainer: Trump’s Week One Actions on AI, What’s Affected, and What Comes Next?
President Trump’s first round of executive orders included a dramatic shift in U.S. AI policy. On his first day in office, Trump repealed the cornerstone of former President Biden’s AI policy, EO ...
Want Insight into AI Policy Under President Trump? Four Questions for Secretary of Commerce Nominee Howard Lutnick
As the Trump Administration makes its first moves on AI policy, the Commerce Department will likely remain a nerve center for AI policymaking over the next four years. In addition to housing NIST and ...
Star-Spangled AI: OpenAI’s Blueprint for Progress and Politics
At the end of December, my colleague Iskandar Haykel and I wrote a blog post recapping three major trends in the AI industry in 2024, one of which was that frontier AI labs are backing away from their...
