Sam Altman outlines three possible paths for safe AI development- Moneycontrol.com

October 23, 2026 / 09:16 IST

OpenAI CEO Sam Altman spoke to employees last week about three ideas the AI industry has discussed to make AI development safer, according to a report by The Information. This comes as the Trump administration is creating its first set of rules to oversee companies developing superintelligence, a form of AI that could become much smarter than humans. Altman reportedly said OpenAI must choose which of these three plans it will support within the next year.

Three suggested plans

The first idea is to create an international organization similar to the International Atomic Energy Agency (IAEA). This organization would make sure AI companies follow safety rules. According to people who knew about Altman's comments, he seemed to prefer this idea. Since 2023, OpenAI leaders have supported creating such an organization. They have suggested that it should inspect AI systems, require safety checks, and keep track of how much computing power and energy companies use.

The second idea is called the Demis plan by Altman. It comes from an earlier suggestion by Demis Hassabis, the head of Google DeepMind. Hassabis suggested creating an organization similar to CERN, where scientists from different countries work together. In Altman's version, major AI companies would join together or work through one organization. This organization would provide AI models to countries that partner with it.

The third idea is called the Eliezer plan, named after AI safety thinker Eliezer Yudkowsky. Under this plan, one AI company would become so powerful financially that it could stop competing companies from getting the computing resources they need. Altman said he was against this idea and mentioned it only to make the list complete. According to the report, some employees thought his comment was a criticism of Anthropic because of Yudkowsky's connections to that company.

AI safety worries and what companies are doing

Altman's comments came after several security breaches involving AI models from OpenAI, Anthropic, Google, and Meta. These were incidents in which AI-related security protections were broken. Altman has told employees privately that OpenAI cannot afford to fall behind its competitors. Both OpenAI and Anthropic are also moving toward selling shares to the public through stock market listings.

At the same time, OpenAI, Google, and Anthropic are working together to create a US-based AI safety group called the Standards Authority for Frontier AI. They hope to launch it by early 2027. Anthropic has also given lawmakers a proposed law that would let the US Commerce Department stop AI models that could cause harm on a very large scale.