Authors of a forecast predicting humanity's extinction due to AI have released a new scenario—this time with a chance for salvation. The document AI 2040: Plan A was presented by former OpenAI researcher Daniel Kokotailo and the AI Futures Project.

Kokotailo left OpenAI in April 2024 due to disagreements over AI safety, and in 2025, he founded the AI Futures Project.

In April of the same year, the organization released the AI 2027 forecast, which stated that the AI race between the U.S. and China could lead to humanity's extinction or dictatorship. The authors noted that the probability of extinction ranges from 10% to 30%.

AI 2040: Plan A is described by experts as a set of recommendations—a series of steps that could mitigate risks.

How It Should Work

According to the scenario, in 2029, the U.S. and China will sign an international agreement to halt the race for superintelligence. Without this deal, AI automation would occur by 2030.

Instead, the countries would gradually develop neural networks to match the best human experts. By 2035, this development would pause to maintain human control over the systems, and by 2040, the pause would be lifted, allowing AI to reach superintelligence.

The plan is based on four principles:

  • buying time for safety research;
  • fully disclosing AI developments;
  • distributing artificial intelligence among various companies and countries;
  • ensuring the reversibility of the process.

To build trust among the parties, Plan A relies on verification. Large data centers are visible from space, making them hard to hide.

This forms the first step: countries publicly declare their purchases of AI chips. Next, they implement a temporary pause on new training runs—sensors at data centers confirm compliance. Once trust is established, the restrictions are lifted, but research remains fully transparent: any country and company can see what others are doing.

To prevent the deal from falling apart, the authors propose "mutually assured destruction of computational power," likening it to the logic of nuclear deterrence. New Chinese data centers are planned to be built in Canada, while U.S. facilities will be in Mongolia, making them easier targets for adversaries in case of conflict.

If the agreement fails, the host country will attempt to seize foreign facilities on its territory, and the owner will destroy them to prevent them from falling into enemy hands.

The Economics of the Plan

According to the authors' calculations, global computing power will grow from 20 million H100 equivalents in 2026 to 60 billion by 2034. The real GDP growth in the U.S. during certain periods of the 2030s could reach 50% per year—compared to the usual 3%.

Due to automation, employment in the United States is expected to drop from 62% in 2027 to 12% by 2040. To compensate for job losses, experts suggest introducing "civil dividends"—payments to every adult American from the revenue the government collects from companies for computing and robot permits.

According to AI Futures Project forecasts, the dividend will amount to $45,000 per person by 2032. By 2035, it will rise to $1 million, and by 2039, it could reach $10 million.

Four Alternative Scenarios

Experts contrasted Plan A with four other potential outcomes. They argue these are storylines of a single scenario—illustrating what might happen under different governmental decisions.

  • Plan B—The U.S. forms a coalition of allies and pressures China, potentially resorting to cyberattacks and targeted strikes on foreign AI projects. This scenario ends with a choice between losing control over AI and war.
  • Plan C—Washington attempts to negotiate with Beijing and implement internal regulations, but under corporate pressure, the pause is quickly lifted. The outcome risks a permanent oligarchy controlling superintelligence.
  • Plan D—Authorities opt for minimal regulation and a race. The risks include losing control over AI, extreme concentration of power, and a third world war.
  • Plan S—A complete indefinite halt to advanced AI developments. The main danger is that the deal will eventually collapse, and the race will resume under less controlled conditions.

It is worth noting that in February, Citrini Research experts predicted an economic collapse due to artificial intelligence.