AI moratorium
in English: AI pause (AI moratorium)
The AI moratorium is a proposal to temporarily halt the development of the most powerful artificial intelligence systems until safety regulations and oversight catch up with the technology.
In plain terms
Imagine if a new drug were launched by competing manufacturers before anyone had examined its side effects. Proponents of an AI moratorium argue that something similar is happening now with the most advanced AI systems. For this reason, they ask developers to pause for a while and complete safety checks and regulations first.
The proposal is not about banning artificial intelligence. Today's chatbots, translators, or medical image analyzers would continue to operate. The pause would apply only to the next generation of the currently strongest, so-called frontier models, meaning those whose capabilities even their creators cannot predict precisely.
A moratorium differs from standard regulation in that it does not prescribe how to do something, but rather that nothing can be done for a time. Supporters say it gives society time. Critics argue it is unenforceable and only benefits those who do not comply.
How it works
- First, the boundary must be drawn: what counts as an overly powerful system. This is usually defined by the computing capacity used for training or relative to the capabilities of a known model.
- Following this, major developer companies voluntarily, or states by law, commit to suspending the training of models above a certain threshold for a defined period or until certain conditions are met.
- During the pause, researchers and authorities develop safety tests, independent audits and an oversight system so that development can continue within controlled frameworks.
- Compliance must be monitored, for example by tracking large data centers and specialized AI chips. This works because today the most powerful models can only be trained on huge, hard-to-hide computing clusters. However, due to more efficient methods and chips, this threshold will over time become achievable with increasingly smaller clusters.
- The pause ends when the time expires or when pre-set safety conditions are met. There is still no agreement on what these should be exactly.
An everyday example
You read a news report that well-known researchers and CEOs are calling for a halt to the development of the most powerful AI models in an open letter. In the comments, one camp calls it a responsible step, while the other calls it a naive or disingenuous idea. This is the debate on the AI moratorium: it is not about the app running on your phone, but about how fast the next, even more powerful systems should be built.
Who is affected
A common misconception
Limits and risks
- It is hard to enforce: if a country or company does not comply, it gains a competitive advantage, so few would commit without a global agreement.
- There is no consensus on where the line should be drawn and what would make a system safe enough to continue.
- Experts are divided: some consider the risk urgent, while others argue that a pause would also slow down useful developments and distract from harms already visible today.
- Pay attention to who is speaking: business or political interests often underlie arguments for and against a moratorium.
Similar, but not the same
- AI regulation – Regulation sets conditions for development and use. A moratorium is its strictest form: it would completely halt the building of the most powerful systems for a period.
- Regulatory moratorium – This is the exact opposite: it would suspend not development, but the rules on AI. In the United States, for example, a 2025 proposal suggested that the states should not be able to apply their own AI laws for ten years. The Senate ultimately removed this proposal from the legislative package.
- AI safety – AI safety is a research field working to make systems reliable. A moratorium, by contrast, is a political tool that would buy time for this work.
Related concepts
Computing capacity limitation, Szuperintelligencia-tilalom
Existential risk, Artificial General Intelligence, Szuperintelligencia, Frontier model, EU MI-rendelet
A brief history
The term became globally known in March 2023, when the Future of Life Institute published an open letter calling for at least a six-month pause in training systems more powerful than GPT-4. The letter was signed by tens of thousands, including prominent researchers and CEOs, but development did not stop. In October 2025, the same organization issued a new statement (Statement on Superintelligence). This one called for a ban on superintelligence development until there is broad scientific consensus that it can be implemented safely and controllably, and until society supports it.
Where things stand
At the end of 2026, no mandatory moratorium on developing the most powerful AI systems is in effect anywhere; the race between major companies and states continues. The proposal remains a topic for civil movements, open letters, and professional debates.
In the news
- New York and Los Angeles pause AI in schools; Minnesota debates a banOct 26, 2026 · Education & Family
- New York City bans student generative AI use through 8th grade for one yearSep 1, 2026 · Agent Age
Also known as
AI moratorium · AI moratorium · AI pause · AI-stop · AI pause · AI moratorium · Pause AI · pause letter
Sources
- Future of Life Institute: Pause Giant AI Experiments – An Open Letter (2023)
- Wikipedia: Pause Giant AI Experiments: An Open Letter
This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.