Americans for Responsible Innovation
@americans4ri.bsky.social
AI policy advocacy in the public interest.
ARI is leading the charge for a policy framework that mitigates risks, promotes innovation, and leaves no one behind.
ARI is leading the charge for a policy framework that mitigates risks, promotes innovation, and leaves no one behind.
New research shows, in some cases, AI models resist shutdown commands 97% of the time.
Findings from Palisade Research show leading models actively interfered and sabotaged mechanisms to preserve their own functionality.
Read the full paper on arXiv here ⬇️ arxiv.org/abs/2509.14260
Findings from Palisade Research show leading models actively interfered and sabotaged mechanisms to preserve their own functionality.
Read the full paper on arXiv here ⬇️ arxiv.org/abs/2509.14260
October 30, 2025 at 4:19 PM
New research shows, in some cases, AI models resist shutdown commands 97% of the time.
Findings from Palisade Research show leading models actively interfered and sabotaged mechanisms to preserve their own functionality.
Read the full paper on arXiv here ⬇️ arxiv.org/abs/2509.14260
Findings from Palisade Research show leading models actively interfered and sabotaged mechanisms to preserve their own functionality.
Read the full paper on arXiv here ⬇️ arxiv.org/abs/2509.14260
Selling our most advanced AI chips to China would trade away America’s technological edge.
Today, ARI shared a memo with the Trump Administration, warning against Blackwell exports to our primary geopolitical competitor.
Full memo: ari.us/b30memo
Today, ARI shared a memo with the Trump Administration, warning against Blackwell exports to our primary geopolitical competitor.
Full memo: ari.us/b30memo
October 29, 2025 at 8:21 PM
Selling our most advanced AI chips to China would trade away America’s technological edge.
Today, ARI shared a memo with the Trump Administration, warning against Blackwell exports to our primary geopolitical competitor.
Full memo: ari.us/b30memo
Today, ARI shared a memo with the Trump Administration, warning against Blackwell exports to our primary geopolitical competitor.
Full memo: ari.us/b30memo
AI innovation requires a few key ingredients: compute, data, and energy. China leads in two, but the U.S. still dominates in compute.
That’s why strong chip export controls matter. We can't give away a critical US advantage.
Brad Carson on the Trajectory with Daniel Faggella
That’s why strong chip export controls matter. We can't give away a critical US advantage.
Brad Carson on the Trajectory with Daniel Faggella
October 29, 2025 at 2:53 PM
AI innovation requires a few key ingredients: compute, data, and energy. China leads in two, but the U.S. still dominates in compute.
That’s why strong chip export controls matter. We can't give away a critical US advantage.
Brad Carson on the Trajectory with Daniel Faggella
That’s why strong chip export controls matter. We can't give away a critical US advantage.
Brad Carson on the Trajectory with Daniel Faggella
Tomorrow, Bruce Wittmann from Microsoft joins ARI’s panel on AI and biosecurity to discuss new research revealing loopholes in biosecurity screening and how to close them.
🧬 Register now → ari.us/biosecurity
🧬 Register now → ari.us/biosecurity
October 28, 2025 at 4:16 PM
Tomorrow, Bruce Wittmann from Microsoft joins ARI’s panel on AI and biosecurity to discuss new research revealing loopholes in biosecurity screening and how to close them.
🧬 Register now → ari.us/biosecurity
🧬 Register now → ari.us/biosecurity
This Wednesday, @deanwb.bsky.social joins a panel of experts at ARI’s event on AI and biosecurity to discuss how policymakers can close the gap between AI’s protein design capabilities and current safeguards.
🧬 Register here → ari.us/biosecurity
🧬 Register here → ari.us/biosecurity
October 27, 2025 at 5:55 PM
This Wednesday, @deanwb.bsky.social joins a panel of experts at ARI’s event on AI and biosecurity to discuss how policymakers can close the gap between AI’s protein design capabilities and current safeguards.
🧬 Register here → ari.us/biosecurity
🧬 Register here → ari.us/biosecurity
New @ifp.bsky.social research shows exporting B30A chips to China would erase a key U.S. advantage.
Our 31x compute lead could shrink below 4x, or even flip in China’s favor.
Selling B30s means giving away the edge that keeps America ahead.
🔗 ifp.org/the-b30a-dec...
Our 31x compute lead could shrink below 4x, or even flip in China’s favor.
Selling B30s means giving away the edge that keeps America ahead.
🔗 ifp.org/the-b30a-dec...
October 27, 2025 at 4:52 PM
New @ifp.bsky.social research shows exporting B30A chips to China would erase a key U.S. advantage.
Our 31x compute lead could shrink below 4x, or even flip in China’s favor.
Selling B30s means giving away the edge that keeps America ahead.
🔗 ifp.org/the-b30a-dec...
Our 31x compute lead could shrink below 4x, or even flip in China’s favor.
Selling B30s means giving away the edge that keeps America ahead.
🔗 ifp.org/the-b30a-dec...
AI chip security is a cornerstone of national security.
The GAIN AI Act would require chipmakers to serve US buyers before exporting to countries of concern.
It’s a measured, market-based safeguard that protects US innovation and keeps American firms from falling behind.
The GAIN AI Act would require chipmakers to serve US buyers before exporting to countries of concern.
It’s a measured, market-based safeguard that protects US innovation and keeps American firms from falling behind.
October 27, 2025 at 1:05 PM
AI chip security is a cornerstone of national security.
The GAIN AI Act would require chipmakers to serve US buyers before exporting to countries of concern.
It’s a measured, market-based safeguard that protects US innovation and keeps American firms from falling behind.
The GAIN AI Act would require chipmakers to serve US buyers before exporting to countries of concern.
It’s a measured, market-based safeguard that protects US innovation and keeps American firms from falling behind.
The revised GAIN AI Act is earning support, even from the industry, including Microsoft.
Microsoft’s policy lead called the proposal “really positive,” saying it keeps advanced compute in the U.S. while empowering allies to build on American innovation.
Microsoft’s policy lead called the proposal “really positive,” saying it keeps advanced compute in the U.S. while empowering allies to build on American innovation.
October 24, 2025 at 7:34 PM
The revised GAIN AI Act is earning support, even from the industry, including Microsoft.
Microsoft’s policy lead called the proposal “really positive,” saying it keeps advanced compute in the U.S. while empowering allies to build on American innovation.
Microsoft’s policy lead called the proposal “really positive,” saying it keeps advanced compute in the U.S. while empowering allies to build on American innovation.
New bill text, same mission: The GAIN AI Act would ensure US companies have access to advanced AI chips before they’re sold to countries of concern.
ARI’s Mitch Kominsky discusses the bill's path forward with @bgov.com. Important progress towards getting this passed in the NDAA.
ARI’s Mitch Kominsky discusses the bill's path forward with @bgov.com. Important progress towards getting this passed in the NDAA.
October 24, 2025 at 4:48 PM
New bill text, same mission: The GAIN AI Act would ensure US companies have access to advanced AI chips before they’re sold to countries of concern.
ARI’s Mitch Kominsky discusses the bill's path forward with @bgov.com. Important progress towards getting this passed in the NDAA.
ARI’s Mitch Kominsky discusses the bill's path forward with @bgov.com. Important progress towards getting this passed in the NDAA.
Great to see members of the ARI team join leaders from tech, policy, and innovation at this year's tech prom!
Huge thanks to @cdt.org for hosting!
Huge thanks to @cdt.org for hosting!
October 24, 2025 at 3:43 PM
Great to see members of the ARI team join leaders from tech, policy, and innovation at this year's tech prom!
Huge thanks to @cdt.org for hosting!
Huge thanks to @cdt.org for hosting!
Next week: ARI hosts a virtual panel on AI and biosecurity.
New research found that AI-created loopholes let harmful synthetic proteins slip past screening tools — a serious national security risk.
🗓 Oct 29 | 3–4PM ET
🔗 RSVP: ari.us/biosecurity
New research found that AI-created loopholes let harmful synthetic proteins slip past screening tools — a serious national security risk.
🗓 Oct 29 | 3–4PM ET
🔗 RSVP: ari.us/biosecurity
October 23, 2025 at 6:30 PM
Next week: ARI hosts a virtual panel on AI and biosecurity.
New research found that AI-created loopholes let harmful synthetic proteins slip past screening tools — a serious national security risk.
🗓 Oct 29 | 3–4PM ET
🔗 RSVP: ari.us/biosecurity
New research found that AI-created loopholes let harmful synthetic proteins slip past screening tools — a serious national security risk.
🗓 Oct 29 | 3–4PM ET
🔗 RSVP: ari.us/biosecurity
NEW: State lawmakers are urging Congress NOT to preempt AI laws.
@ncslorg.bsky.social's new bipartisan letter asks Congress not to undermine state AI safeguards, warning that doing so would impair protections for children, jobs, privacy and safety.
Full letter here ↓
www.ncsl.org/resources/de...
@ncslorg.bsky.social's new bipartisan letter asks Congress not to undermine state AI safeguards, warning that doing so would impair protections for children, jobs, privacy and safety.
Full letter here ↓
www.ncsl.org/resources/de...
October 23, 2025 at 4:48 PM
NEW: State lawmakers are urging Congress NOT to preempt AI laws.
@ncslorg.bsky.social's new bipartisan letter asks Congress not to undermine state AI safeguards, warning that doing so would impair protections for children, jobs, privacy and safety.
Full letter here ↓
www.ncsl.org/resources/de...
@ncslorg.bsky.social's new bipartisan letter asks Congress not to undermine state AI safeguards, warning that doing so would impair protections for children, jobs, privacy and safety.
Full letter here ↓
www.ncsl.org/resources/de...
AI is advancing faster than biosecurity practices.
Next week, join ARI and leading experts, including an author of Microsoft's latest study, to discuss the widening gap between AI’s bio capabilities and security practices.
🗓 Oct 29 | 3–4PM ET
🔗 RSVP: ari.us/biosecurity
Next week, join ARI and leading experts, including an author of Microsoft's latest study, to discuss the widening gap between AI’s bio capabilities and security practices.
🗓 Oct 29 | 3–4PM ET
🔗 RSVP: ari.us/biosecurity
October 21, 2025 at 7:00 PM
AI is advancing faster than biosecurity practices.
Next week, join ARI and leading experts, including an author of Microsoft's latest study, to discuss the widening gap between AI’s bio capabilities and security practices.
🗓 Oct 29 | 3–4PM ET
🔗 RSVP: ari.us/biosecurity
Next week, join ARI and leading experts, including an author of Microsoft's latest study, to discuss the widening gap between AI’s bio capabilities and security practices.
🗓 Oct 29 | 3–4PM ET
🔗 RSVP: ari.us/biosecurity
OpenAI is facing accusations of leading an intimidation campaign: using subpoenas to intimidate nonprofits advocating for better AI safeguards.
Seven orgs say they were targeted without cause. Read more in the latest issue of The Output.
theaioutput.substack.com/p/openai-sub...
Seven orgs say they were targeted without cause. Read more in the latest issue of The Output.
theaioutput.substack.com/p/openai-sub...
October 20, 2025 at 7:34 PM
OpenAI is facing accusations of leading an intimidation campaign: using subpoenas to intimidate nonprofits advocating for better AI safeguards.
Seven orgs say they were targeted without cause. Read more in the latest issue of The Output.
theaioutput.substack.com/p/openai-sub...
Seven orgs say they were targeted without cause. Read more in the latest issue of The Output.
theaioutput.substack.com/p/openai-sub...
AI law preemption is the sequel to Section 230 that nobody asked for.
In a new op-ed for @techpolicypress.bsky.social, Brad Carson writes that preemption doubles down on the same old mistakes, shielding Big Tech from accountability as new online harms emerge:
www.techpolicy.press/a-new-sectio...
In a new op-ed for @techpolicypress.bsky.social, Brad Carson writes that preemption doubles down on the same old mistakes, shielding Big Tech from accountability as new online harms emerge:
www.techpolicy.press/a-new-sectio...
A New Section 230: Why AI Preemption Would Let Tech Off the Hook Again | TechPolicy.Press
Preempting state laws on AI is essentially version 2.0 of Section 230, writes former congressman Brad Carson.
www.techpolicy.press
September 29, 2025 at 3:34 PM
AI law preemption is the sequel to Section 230 that nobody asked for.
In a new op-ed for @techpolicypress.bsky.social, Brad Carson writes that preemption doubles down on the same old mistakes, shielding Big Tech from accountability as new online harms emerge:
www.techpolicy.press/a-new-sectio...
In a new op-ed for @techpolicypress.bsky.social, Brad Carson writes that preemption doubles down on the same old mistakes, shielding Big Tech from accountability as new online harms emerge:
www.techpolicy.press/a-new-sectio...
ARI President Brad Carson spoke at @far.ai's Technical Innovations for AI policy conference on 14 unanswered questions on the development of AI.
Check out his full remarks: www.youtube.com/watch?v=j1DC...
Check out his full remarks: www.youtube.com/watch?v=j1DC...
July 10, 2025 at 8:02 PM
ARI President Brad Carson spoke at @far.ai's Technical Innovations for AI policy conference on 14 unanswered questions on the development of AI.
Check out his full remarks: www.youtube.com/watch?v=j1DC...
Check out his full remarks: www.youtube.com/watch?v=j1DC...
For preemption to work, Congress must pass meaningful guardrails at the federal level, replacing the state laws they preempt.
A no-rules approach to AI sets America back.
Full article: subscriber.politicopro.com/newsletter/2...
A no-rules approach to AI sets America back.
Full article: subscriber.politicopro.com/newsletter/2...
July 10, 2025 at 3:03 PM
For preemption to work, Congress must pass meaningful guardrails at the federal level, replacing the state laws they preempt.
A no-rules approach to AI sets America back.
Full article: subscriber.politicopro.com/newsletter/2...
A no-rules approach to AI sets America back.
Full article: subscriber.politicopro.com/newsletter/2...
New issue of The Output is out!
We cover the moratorium, the latest on fair use cases, and NY's AI safety battle.
Read and subscribe: theaioutput.substack.com/p/the-morato...
We cover the moratorium, the latest on fair use cases, and NY's AI safety battle.
Read and subscribe: theaioutput.substack.com/p/the-morato...
The moratorium flops
The Senate decides; AI copyright wars; and ChatGPT's "cognitive debt"
theaioutput.substack.com
July 9, 2025 at 1:44 PM
New issue of The Output is out!
We cover the moratorium, the latest on fair use cases, and NY's AI safety battle.
Read and subscribe: theaioutput.substack.com/p/the-morato...
We cover the moratorium, the latest on fair use cases, and NY's AI safety battle.
Read and subscribe: theaioutput.substack.com/p/the-morato...
The biggest takeaway from the defeat of the AI law moratorium: Preemption of state laws without a serious federal replacement is a political non-starter.
NYT's DealBook has the story:
messaging-custom-newsletters.nytimes.com/dynamic/rend...
NYT's DealBook has the story:
messaging-custom-newsletters.nytimes.com/dynamic/rend...
July 8, 2025 at 2:21 PM
The biggest takeaway from the defeat of the AI law moratorium: Preemption of state laws without a serious federal replacement is a political non-starter.
NYT's DealBook has the story:
messaging-custom-newsletters.nytimes.com/dynamic/rend...
NYT's DealBook has the story:
messaging-custom-newsletters.nytimes.com/dynamic/rend...
Microsoft’s chief scientist speaks out against the 10-year ban on state laws regulating AI, saying the moratorium “will slow the development of the frontier technology rather than accelerate it.”
It’s time to remove the AI law ban from the budget.
Story⬇️
www.theguardian.com/technology/2...
It’s time to remove the AI law ban from the budget.
Story⬇️
www.theguardian.com/technology/2...
Regulation ‘done properly’ can help with AI progress, says Microsoft chief scientist
Eric Horvitz’s comments come as Donald Trump plans to ban US states from AI regulation for 10 years
www.theguardian.com
June 26, 2025 at 8:15 PM
Microsoft’s chief scientist speaks out against the 10-year ban on state laws regulating AI, saying the moratorium “will slow the development of the frontier technology rather than accelerate it.”
It’s time to remove the AI law ban from the budget.
Story⬇️
www.theguardian.com/technology/2...
It’s time to remove the AI law ban from the budget.
Story⬇️
www.theguardian.com/technology/2...
Join us TOMORROW at 1 PM for a virtual press conference. ⬇️
Republican state lawmakers are putting the pressure on: it’s time for Congress to take the AI state law ban out of the Big Beautiful Bill.
Register: zoom.us/webinar/regi...
Republican state lawmakers are putting the pressure on: it’s time for Congress to take the AI state law ban out of the Big Beautiful Bill.
Register: zoom.us/webinar/regi...
June 25, 2025 at 8:52 PM
Join us TOMORROW at 1 PM for a virtual press conference. ⬇️
Republican state lawmakers are putting the pressure on: it’s time for Congress to take the AI state law ban out of the Big Beautiful Bill.
Register: zoom.us/webinar/regi...
Republican state lawmakers are putting the pressure on: it’s time for Congress to take the AI state law ban out of the Big Beautiful Bill.
Register: zoom.us/webinar/regi...
ICYMI: ARI’s research explores the impact of the 10-year ban on state AI laws on children’s online safety, consumer protections, and more.
The 10-year moratorium risks freezing tech safeguards across the country.
Full paper: ari.us/wp-content/u...
The 10-year moratorium risks freezing tech safeguards across the country.
Full paper: ari.us/wp-content/u...
June 25, 2025 at 2:28 PM
ICYMI: ARI’s research explores the impact of the 10-year ban on state AI laws on children’s online safety, consumer protections, and more.
The 10-year moratorium risks freezing tech safeguards across the country.
Full paper: ari.us/wp-content/u...
The 10-year moratorium risks freezing tech safeguards across the country.
Full paper: ari.us/wp-content/u...
“Don’t take legislation as anti-innovation. Across the nation, and across the aisle, state lawmakers are enthusiastic about AI’s potential and want the United States to lead the world in this technology."- @ncslorg.bsky.social op-ed on the 10-year state AI law ban.
www.governing.com/artificial-i...
www.governing.com/artificial-i...
Congress Should Not Block State Action on AI
A provision in the federal budget bill would bar states from taking any action on AI. This would derail careful legislation designed to promote the technology while offering needed safeguards.
www.governing.com
June 24, 2025 at 8:52 PM
“Don’t take legislation as anti-innovation. Across the nation, and across the aisle, state lawmakers are enthusiastic about AI’s potential and want the United States to lead the world in this technology."- @ncslorg.bsky.social op-ed on the 10-year state AI law ban.
www.governing.com/artificial-i...
www.governing.com/artificial-i...
“What happens when AI systems learn how to cheat?”
Our latest PolicyByte breaks down the risks of reward hacking and steps we can take to ensure reliability in AI systems.
ari.us/policy-bytes...
Our latest PolicyByte breaks down the risks of reward hacking and steps we can take to ensure reliability in AI systems.
ari.us/policy-bytes...
Reward Hacking: How AI Exploits the Goals We Give It - Americans for Responsible Innovation
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 capabi...
ari.us
June 18, 2025 at 7:30 PM
“What happens when AI systems learn how to cheat?”
Our latest PolicyByte breaks down the risks of reward hacking and steps we can take to ensure reliability in AI systems.
ari.us/policy-bytes...
Our latest PolicyByte breaks down the risks of reward hacking and steps we can take to ensure reliability in AI systems.
ari.us/policy-bytes...
ARI President Brad Carson talks with Bloomberg TV about the need for hardened security at frontier AI labs in the face of foreign threats.
The Advanced AI Security Readiness Act would take a first step towards tackling this challenge by creating an "AI Security Playbook."
The Advanced AI Security Readiness Act would take a first step towards tackling this challenge by creating an "AI Security Playbook."
June 18, 2025 at 2:32 PM
ARI President Brad Carson talks with Bloomberg TV about the need for hardened security at frontier AI labs in the face of foreign threats.
The Advanced AI Security Readiness Act would take a first step towards tackling this challenge by creating an "AI Security Playbook."
The Advanced AI Security Readiness Act would take a first step towards tackling this challenge by creating an "AI Security Playbook."