Our Priorities
We believe AI should be built to serve humanity: developed and deployed under meaningful human control, with clear accountability when it causes harm. Safeguards do not stifle innovation; they are necessary to build the public trust required to ensure this technology reaches its full potential.
The Alliance for Secure AI works to build enforceable safeguards in these key areas:
Prove It's Safe Before We Use It: Safety Testing and Model Evaluation
No other powerful technology — not cars, not medicines, not aircraft — reaches the public without being tested first. Advanced AI is the exception. There is still no national baseline requirement to show that the most capable systems behave as intended before they are released.
We need clear standards for evaluating advanced models before and after deployment, independent scrutiny of dangerous capabilities, and meaningful human control over consequential decisions. This is not about slowing innovation; it is about ensuring America leads in building both cutting-edge AI and the methods that prove it works as promised. Legislation like California’s SB 53, New York’s RAISE Act, Illinois’s SB 315, and the federal Artificial Intelligence Risk Evaluation Act reflect this approach.
When AI Causes Harm, Someone Is Responsible: Accountability and Liability
AI is already screening job applicants, advising patients, and shaping financial decisions. When it gets those calls wrong, the people harmed should have clear recourse, and the companies that built the system should be held responsible.
The law should reflect the simple fact that AI is a product, not a person. Our long-standing product-liability standards should apply to AI products. Developers should answer for foreseeable harms and be required to build strong protections for children and other vulnerable users. Just as important is what the law should not do: grant blanket immunity that would leave families without recourse when these systems cause real harm. AI is a powerful tool and it should never become a way to escape responsibility. Federal bills like the AI LEAD Act and the GUARD Act create the right model.
Workers and Communities Shouldn't Bear the Cost Alone: Societal and Environmental Impact
AI is already reshaping who works and how, automating skilled tasks and thinning the entry-level jobs that have long been a path into a career. As these systems grow more capable, that pressure will only build.
The infrastructure behind AI carries its own costs. Training and running advanced models demands enormous energy and resources, straining local grids and driving up electricity bills. Policymakers should weigh those long-term costs honestly, not just the short-term growth claims. Americans are told AI would expand opportunity; policy should make sure that promise reaches workers and communities, not just balance sheets and stock prices.
AI Should Expand Freedom, Not Surveillance: Privacy and Civil Liberties
AI is making surveillance cheaper, faster, and nearly unlimited. License-plate readers log where we drive; data brokers package our personal lives for sale; facial recognition can pick a face out of a crowd. These tools can serve legitimate public safety needs, but without limits they hand government and corporations alike an unprecedented window into the lives of ordinary Americans.
Sensible limits keep AI a tool for people, not a tool turned against them: closing the data-broker loophole that lets government agencies buy personal information they would otherwise need a warrant to obtain, real transparency and safeguards around automated surveillance like license-plate readers, and meaningful oversight of how these systems are deployed. Privacy and security are not opposing goals, and a free society should protect both.
Win the AI Race Without Losing Control of the Technology: National Security and U.S. Leadership
AI is now a front-line national security asset. The same advanced chips and models that drive American innovation can, in the wrong hands, accelerate cyberattacks, autonomous weapons, and the surveillance regimes of authoritarian rivals. Protecting that advantage is inseparable from using it responsibly.
Maintaining American leadership means keeping our most advanced technology out of adversaries’ hands: enforcing export controls, closing the loopholes that let restricted chips slip through, and securing supply chains against theft and espionage. Real leadership is not only building the most capable AI in the world. It is proving, to allies and citizens alike, that it can be trusted. Bills like the MATCH Act, the Chip Security Act, and the No Adversarial AI Act are key priorities to ensure that outcome.