The Humane AI Manifesto

Humane By choice.

AI should serve the people who will live with its consequences.

Our mission

Our mission is to advance more humane ways of implementing, governing, and using AI by putting shared values and principles into practice and learning through collaboration.

The manifesto

Four principles.

  1. Continuous learning
  2. Long-term sustainability
  3. Real-world evidence
  4. Adaptive governance

Seven values.

Human Dignity Over Capability

Human dignity begins with every person’s inherent worth and inalienable rights. AI should support our ability to contribute to society and improve human decisions, workflows, safety, creativity, and productivity. It should add to life, not detract from it. As AI becomes more capable and autonomous, human dignity and humanity’s continued existence must remain at the center of how we develop, use, and deploy it. The future we are building should serve the people who will live in it. Automation should amplify expert capability, not absolve accountability. The highest-value AI systems increase capability, reduce cognitive load, surface insight, and accelerate learning.

Human Agency Over Scale

Require human oversight at every stage, with no siloed releases and clear ownership of outcomes. Prioritize informed consent and human agency over speed and convenience. People must understand when, where, and how AI is used, and who is accountable when systems fail or cause harm. Designs that significantly compromise lives and livelihoods or exploit disenfranchised communities are unacceptable. Assess and address environmental impacts, mass layoffs, strain on unemployment systems, and the loss of opportunities to develop future operators, implementers, and thought leaders.

Human Judgement Over Automation

The strongest long-term organizations combine human adaptability with scalability. Fully autonomous lethal weapons bring these stakes into sharp focus. Without meaningful human control, decisions about taking a life could rest with algorithms whose biases and decision-making processes we may not fully understand. We risk giving AI power over human lives when it cannot bear the consequences of those decisions. Decisions with such irreversible consequences demand human judgment, oversight, and accountability.

Long-term Sustainability Over Short-term Expediency

Robust solutions that consistently produce safe, reliable outcomes for humans and our environment conserve human time and fragile ecosystems. Quality over quantity. Design for resiliency, including model drift, outages, adversarial prompts, dependency failures, and changing regulations. Lean over waste, such as verbose code, inefficient token generation (by prompts) or profligate token generation (by agents).

Adaptive Policies Over Rigid Control

Governance must evolve with model capabilities, attack patterns, and organizational maturity. Heavy governance that blocks learning creates shadow AI use and hidden risks. AI systems are never finished/complete. Models, prompts, retrieval systems, user behavior, and policies continually evolve. Policies must be relative to maturity and potential consequence.

Transparent Limitations Over Superficial Intelligence

AI systems need to clearly communicate uncertainty, confidence, boundaries, limitations (including data limitations), and escalation paths. Trust builds faster from honest imperfection than simulated omniscience. Let imperfections be visible.

Evidence of Safety Over Assurances of Safety

Demonstrate safety through empirical evaluation, continuous learning from real world use, clear boundaries, and continuous oversight, not just corporate pledges and codes of conduct. Responsible deployment demands tangible verification tools align with intended design, evidence based accuracy, risks rigorously mitigated, and protections adapted alongside changing capabilities.

Founding signatories · 2026

Add your
name.

Stand behind a more humane future with AI.

The choices we make with AI will shape our workplaces, communities, and opportunities for generations. Sign as an individual or as an authorized representative of your organization.

Individuals and organizations that sign during 2026 will be recognized as founding signatories of the Humane AI Manifesto, following verification.

Signing will open when email confirmation is ready. You can preview both signing options below.

How are you signing?

Your email stays private. An affiliation identifies you in a personal capacity. Privacy details

In support of the manifesto

Signatories.

Loading signatories…

Why we
started.

What should we automate? Who remains accountable? How do we protect people whose lives and livelihoods are affected?

The Humane AI Manifesto grew from these questions. Inspired by the Agile Manifesto’s use of shared values to guide decisions, it addresses how AI is implemented, governed, and used, and what those choices mean for the people affected by them.

These values are for anyone who builds, adopts, governs, or uses AI, and for the people whose lives are shaped by it.