Frequently asked questions

Humanity Decides is a public campaign calling for democratic control over the development of artificial superintelligence.

We believe AI can provide enormous benefits. But creating systems more intelligent and capable than humanity as a whole raises a different class of risk.

Decisions of that magnitude should not be made by a small number of companies, researchers, investors, or governments acting independently. Humanity should have a say in whether, when, and under what conditions such systems are created.

No.

AI is already useful across science, medicine, education, engineering, business, accessibility, and many other fields. We expect AI to remain an important part of modern life.

Our concern is specifically the race to create artificial superintelligence: systems capable of exceeding human capabilities across most or all important cognitive domains.

Using AI is one thing. Creating an intelligence more powerful than humanity is another.

No.

Humanity Decides is not calling for an end to ordinary AI research, AI products, medical AI, scientific tools, productivity software, or other beneficial applications.

The focus is on development that could lead to artificial superintelligence or other systems whose capabilities could exceed our ability to reliably understand, supervise, or control them.

No.

The question of whether humanity should retain control over systems more intelligent than itself is not inherently left-wing or right-wing.

People can disagree about taxes, regulation, government, culture, and almost every other political issue while still believing that decisions about superintelligence deserve democratic oversight.

Humanity Decides is non-partisan and funded by private individuals who care about the future of humanity.

It is not for profit, not for sale, and never will be. No participants have any direct financial interest in the outcomes of AI, including private or public holdings in AI labs, regulators, or companies.

No. Humanity Decides is not funded or controlled by an AI laboratory or technology company.

Yes.

AI adoption is unavoidable, and it is here to stay. It can provide enormous benefits to humanity.

But the creation of a superintelligence more powerful than humanity is a different matter. Its development should be regulated until we can establish a credible path to ensuring that such systems remain peacefully aligned with human interests.

Using AI to help write software, create graphics, analyze information, or improve communication is entirely consistent with that position.

This is not an anti-AI website.

The distinction is between using AI as a tool and creating systems powerful enough to exceed human control.

Using electricity does not require supporting every possible application of nuclear technology. Using biotechnology does not require opposing regulations on dangerous pathogens.

Technologies can be useful while particular applications still warrant strict controls.

Superintelligence generally refers to an artificial intelligence that substantially exceeds human cognitive capabilities across a broad range of important tasks.

That could include scientific research, software development, strategic planning, persuasion, engineering, technological invention, and other areas.

There is no universally agreed threshold at which an AI becomes “superintelligent.” That uncertainty is part of the problem. We should establish clear standards before reaching that point, rather than arguing about definitions after it has happened.

A sufficiently capable AI system could potentially make plans, develop technologies, write software, manipulate information, operate through computer systems, or pursue objectives at a speed and scale far beyond human institutions.

The central problem is control.

We do not yet know how to guarantee that an AI more capable than its creators will reliably pursue human intentions, respect human constraints, and remain under meaningful human control.

The consequences of getting that wrong could be severe.

No one knows whether artificial superintelligence will be created, exactly when it might happen, or what its consequences would be.

But the possibility is taken seriously by leading AI researchers, executives, computer scientists, and governments.

The argument for precaution does not require certainty that catastrophe will occur. It requires recognizing that sufficiently powerful systems could create consequences that are difficult or impossible to reverse.

Today’s systems have limitations.

They also improve rapidly, as the pace of progress and failures of control like the Hugging Face incident show.

The important policy question is not whether today’s chatbot is superintelligent. It is whether society should wait until systems approach dangerous capability levels before establishing rules for what happens next.

Humanity Decides argues that the rules should come first.

Alignment is the problem of making an AI system reliably pursue goals and behaviors that remain compatible with human intentions and interests.

For highly capable systems, this becomes difficult because a system may encounter situations its designers never anticipated.

A solution would need to work not only in laboratory demonstrations, but under real-world conditions and against systems potentially more capable than the humans supervising them.

It’s possible, but we should not build public safety policy around that assumption.

A sufficiently capable system could potentially operate across networks, copy information, automate tasks, influence humans, exploit software vulnerabilities, and evade or anticipate attempts to restrict it.

The more capable the system becomes, the less reasonable it is to assume that a simple off switch solves the control problem.

That possibility is one of the strongest arguments for advanced AI.

A safely controlled superintelligence could conceivably produce extraordinary benefits.

But potential benefits do not eliminate the need to understand the risks.

The goal should be to obtain the benefits of advanced intelligence without creating a system whose behavior humanity cannot reliably control.

Yes. No technology is risk-free.

The question is how much risk, imposed on whom, with what potential consequences, and whether those affected have any meaningful say.

Humanity Decides is based on the view that creating an intelligence more powerful than humanity deserves an unusually high standard of evidence and public consent.

No one knows.

Experts disagree widely about the probability.

Humanity Decides does not depend on claiming a precise percentage. The case for action rests on the combination of potentially extreme consequences, rapidly advancing capabilities, unresolved control problems, and substantial uncertainty.

When the downside could affect everyone, uncertainty is not the same thing as safety.

Humanity Decides supports binding regulation preventing the uncontrolled creation of artificial superintelligence until there is a credible, independently verifiable basis for believing such systems can remain under meaningful human control.

The exact regulatory framework should be developed transparently with input from scientists, governments, civil society, industry, security experts, and the public.

The principle is simple: capability should not outrun control.

No.

The goal is not to prohibit advanced intelligence forever.

The goal is to avoid creating systems more powerful than humanity before we know how to control them safely.

If reliable alignment, control, governance, and verification mechanisms are developed, the question can be reconsidered.

Because regulation is most useful before a dangerous threshold is crossed.

Society regulates nuclear materials, pharmaceuticals, aircraft, biological research, financial institutions, and other high-consequence technologies before failure occurs.

If superintelligence could create unusually large or irreversible risks, waiting for definitive evidence after deployment may be a poor safety strategy.

Because governance usually moves more slowly than technology.

Laws, treaties, technical standards, verification systems, and international institutions take years to build.

If governments wait until superintelligence is imminent before beginning that work, they may have very little time left to act.

That question should be answered through law, technical measurement, independent evaluation, and democratic institutions.

It should not be determined solely by the companies building the systems.

Capability thresholds could incorporate factors such as autonomous operation, cyber capabilities, biological research capabilities, replication, strategic planning, persuasion, scientific discovery, and the ability to meaningfully evade human oversight.

The standards will need to evolve as the technology evolves.

Companies have an important role in safety research.

But companies developing frontier AI also face competitive and financial pressure to build more capable systems.

When an activity could impose risks on people who never agreed to take those risks, voluntary corporate policies are not enough.

Independent rules and oversight are standard in other high-consequence industries. AI should not be an exception.

Yes.

Many AI laboratories employ researchers working on alignment, interpretability, evaluations, security, and other safety problems.

That work is valuable.

The question is whether safety research is sufficiently mature to justify creating systems whose capabilities may exceed our ability to control them.

Humanity Decides believes safety should be demonstrated before that threshold is crossed.

It should not.

Most AI innovation does not require building an uncontrolled superintelligence.

There is enormous room for progress in medicine, robotics, science, education, productivity, accessibility, entertainment, engineering, and other fields without crossing every possible capability threshold as quickly as technically possible.

Humanity can continue developing AI while placing limits on the most consequential forms of development.

That is a legitimate concern.

A durable solution will ultimately require international coordination, particularly among countries with advanced AI capabilities.

But competition between countries is not an argument for ignoring a shared risk. Nuclear weapons, biological weapons, aviation, shipping, telecommunications, and other technologies have all required international agreements and verification systems.

The difficulty of cooperation makes the problem harder. It does not make cooperation unnecessary.

Any serious framework would need monitoring, verification, export controls, compute governance, international agreements, and consequences for violations. Nuclear non-proliferation treaties are a useful guide.

No system will be perfect.

The relevant comparison is not between perfect regulation and perfect compliance. It is between attempting to control a potentially dangerous technological race and allowing that race to proceed without meaningful limits.

That is also a risk.

Humanity Decides does not argue that governments should control all AI.

Any regulatory system should contain limits on state power, transparency requirements, independent oversight, due process, and protection for ordinary research and use.

The objective is narrow: prevent uncontrolled development of systems capable of exceeding human control.

Your signature adds your name to a public demand that governments establish meaningful safeguards before artificial superintelligence is developed.

We email you a link to confirm that one real person signed once. Your signature is counted only after you open it.

The purpose of the declaration is to demonstrate politically that this issue extends beyond researchers and technology companies, and that the public expects a voice in decisions of this magnitude.

Only in the form you choose. When you sign, you pick how your name appears: your full name, your first name and last initial, your initials, or kept private. We show you exactly how it will look before you submit.

A private signature still counts in every total, but your name is never listed.

We never publish your email address, postal code, phone number, IP address or your answer to the optional voting question.

We use what you give us only to verify and count your signature, to show your name in the form you chose, and, if you opt in, to send you campaign updates.

We never sell or rent your personal data. The full details are in our Privacy Policy.

Yes. This is a global effort.

Artificial superintelligence would not be a national issue.

People around the world should be able to express their position, although specific declarations or government submissions may distinguish signatures by country or jurisdiction.

Young people have an unusually large stake in decisions whose consequences could extend decades into the future.

You must be at least 13 years old to add your name. If you are under 18, please ask a parent or guardian before you sign.

Anyone, of any age, can share the campaign and talk about why it matters.

A single signature does not determine public policy.

Large numbers of verified signatures can demonstrate public concern, support advocacy, strengthen legislative efforts, attract media attention, and give policymakers evidence that voters care about an issue.

The declaration is a tool for collective political pressure, not an end in itself.

Technical experts should inform the decision.

They should not be the only people allowed to make it.

Scientists can help determine what systems are capable of, what risks exist, and what safety measures may work. But deciding what level of civilization-scale risk society should accept is ultimately a public policy question.

Expertise should inform democratic decisions, not replace them.

No one can say with certainty.

AI development may proceed safely. Technical alignment breakthroughs may occur. Governments or companies may impose effective safeguards voluntarily.

It is also possible that competitive pressure continues driving increasingly capable systems faster than society develops reliable methods to control them.

Humanity Decides exists because we do not believe the second possibility should simply be accepted by default.

Humanity should remain in control of humanity’s future.

Before anyone creates an intelligence more powerful than ourselves, the burden should be on its creators to demonstrate that it can be controlled safely.

And the decision to proceed should belong to society, not merely to whoever gets there first.