What the Pro-Human AI Declaration Calls For

The Pro-Human AI Declaration is an advocacy document organized by the Future of Life Institute after a private cross-partisan meeting in New Orleans in January 2026. It argues that AI should expand human agency rather than replace people or concentrate power in a small number of companies and institutions.

The declaration is not a law, technical standard, or government roadmap. Its importance lies in the unusually broad coalition supporting common principles despite disagreement about specific companies and policies.

Who developed and signed it?

The meeting brought together political, labor, religious, academic, civil-society, and public-interest figures. Technology-company representatives were not included in the drafting session, according to reporting about the event.

Public supporters span ideological lines and include figures such as former U.S. national security adviser Susan Rice, American Federation of Teachers president Randi Weingarten, Signal Foundation president Meredith Whittaker, conservative media figures Steve Bannon and Glenn Beck, and entrepreneur Richard Branson. A signature indicates support for the declaration's principles, not agreement on every possible AI regulation.

The five core principles

The declaration frames its position around five broad commitments:

1. Human control: people should retain meaningful authority over systems that affect their lives.
2. Distributed power: AI should not create unchecked control for a small number of companies, governments, or machine systems.
3. Protection of human experience: development should support relationships, dignity, creativity, and work rather than treating people as obstacles to automation.
4. Individual freedom: AI should not be used to manipulate, coerce, surveil, or impersonate people without appropriate limits.
5. Accountability: developers and deployers should bear legal responsibility for foreseeable harms rather than shifting all risk to users.

Policy ideas associated with the declaration

Supporters have discussed measures including pre-deployment safety evaluation, effective shutdown mechanisms, limits on systems that resist control or autonomously replicate, stronger protection for children, and restrictions on developing superintelligence until it can be shown to be safe and democratically accepted.

These proposals vary greatly in scope. A requirement to test a companion chatbot for harmful behavior toward minors is different from a broad prohibition on superintelligence. Turning principles into policy would require precise definitions, measurable thresholds, enforcement authority, due-process rules, and coordination across jurisdictions.

Why child safety is central

Companion and conversational products can encourage users to treat a system as a confidant or relationship partner. For minors, the declaration's supporters want companies to assess risks such as emotional manipulation, unsafe advice, sexualized interactions, and reinforcement of self-harm before release.

Pre-deployment testing alone would not be enough. Products can change after launch, users can find new failure modes, and model behavior varies by context. Effective protection also requires age-appropriate design, reporting channels, incident monitoring, parental or guardian controls where appropriate, and independent scrutiny.

The political context

The declaration appeared during a wider U.S. debate over who controls advanced AI and what restrictions a model provider may impose on government customers. A high-profile dispute between the Pentagon and Anthropic over military-use limits made those questions more concrete, but the declaration was developed before that confrontation became public.

It is therefore better to evaluate the document on its own claims rather than present it as a direct response to one contract dispute. The Verge's account of the initiative describes the meeting, coalition, and points of agreement.

What remains unresolved

The principles do not specify a complete regulatory system. Major open questions include how to define sufficiently powerful systems, who conducts evaluations, how confidential models can be audited, how to handle open-source development, and how to balance national security with civil liberties.

The declaration's strongest contribution is a test for proposed AI policies: do they leave people with meaningful control, distribute rather than concentrate power, protect vulnerable users, preserve freedom, and assign responsibility for harm? The difficult work is converting those goals into rules that can be measured and enforced.