The Bletchley Declaration (1 November 2023), the G7 Hiroshima Guiding Principles and Code of Conduct (2023), the Seoul Frontier AI Safety Commitments (May 2024), and the OECD AI principles are all soft law. In international law the terms are precise. Hard law means legally binding obligations, chiefly treaties, that states must implement and can be held accountable for breaching. Soft law means instruments that express commitments or standards without creating those obligations: declarations, guiding principles, codes of conduct, resolutions. Both shape behavior. Only hard law can be enforced.

That gap is the difference between soft law and hard law.

What each type can and cannot do

1

Soft law: fast, flexible, non-binding

It can be agreed quickly, adjusted easily, and joined without the full political cost of binding sovereignty. It sets norms, signals intent, and builds shared expectations. It imposes no legal duty, carries no verification, and offers no enforcement. Compliance is voluntary and reversible.

2

Hard law: slow, rigid, binding

A treaty is difficult and slow to negotiate and ratify, and hard to change once agreed. It creates genuine legal obligations, can be paired with verification bodies and enforcement mechanisms, and gives states a basis to hold each other accountable. It is the form that can constrain a determined actor.

Why soft law comes first

The move from soft to hard law is the normal life cycle of international regimes, not a detour. States resist binding constraints on a technology they do not fully understand, where the science is unsettled and the strategic stakes are still contested. Soft law lets them start cooperating without that full commitment, building shared definitions, understanding, and trust that make a binding agreement negotiable later.

History is blunt about the sequence. The Montreal Protocol's binding ozone controls sat on the earlier, non-binding cooperation of the 1985 Vienna Convention. Climate governance began with a framework convention before any binding targets. Soft law is scaffolding. Scaffolding is how structures get built; it is not the finished building.

Soft law also does real work on its own. Norms shape what companies and governments treat as acceptable, influence domestic legislation, and give courts, regulators, and publics reference points they can invoke. The Hiroshima principles and Seoul commitments already change how developers document and describe their systems. None of that is empty.

Why soft law is not enough for frontier AI

  • Voluntary means reversible. A commitment a company or state can abandon under competitive pressure provides no security against a race to cross a dangerous line.
  • No verification means no trust. Unverified commitments leave every party guessing whether others comply. That is fatal in a domain defined by mutual suspicion.
  • No enforcement means no cost to defect. When breaking a promise costs only reputation, the actor who believes the prize is decisive will break it.
  • The stakes can be irreversible. For catastrophic, unrecoverable risks, "we agreed to try our best" is not a safety margin. Binding, verified obligation is.

The folk objection is that hard law is too slow for a fast technology, so soft law must be the permanent answer. Concede the speed problem. Then notice the trade: permanent soft law is a decision to hope that reputation holds when the prize is decisive. Hope is not a control system.

The transition is the task

Soft law is the necessary first stage, and it is doing useful work now. ASI governance cannot stop there. Declarations and codes are valuable as steps toward binding obligation. Their value collapses if the transition never happens.

The central governance task of this decade is to convert soft-law scaffolding into hard-law structure: acknowledged risks into defined thresholds, voluntary commitments into legal duties, cooperative dialogue into verification and enforcement. Every declaration is a down payment. Whether it buys anything depends on whether the world follows through to a treaty before the technology it describes advances beyond reach. See the Foundation's case for preventive international law on superintelligence. Related: the framework-convention path and why the first AI treaty still falls short of frontier catastrophic risk.