AI safety advocates want OpenAI and Anthropic to limit computing power

AI safety advocates are pushing OpenAI and Anthropic to put a limit on the computing power they use to build new AI. The idea is to slow things down so safety measures can keep up with how capable the technology is getting. The report comes from The Information, published October 21, 2026.

What is being proposed

Computing power is the raw processing muscle, mostly in large data centers, that labs use to train AI systems. A cap would set a ceiling on how much of it a lab can use for research and development. Advocates see it as a way to "pace" progress, according to the report.

Altman's three scenarios

Last week, OpenAI CEO Sam Altman laid out three scenarios he said have been considered in the industry for slowing AI down. One of them is merging all of the AI labs into a single entity that would develop new AI safely. The material available does not spell out the other two.

Who it affects

Most people will never train an AI system. But the pace of improvement decides how quickly the chatbots and business tools you use get better, and how soon new ones arrive. A slower pace could change what your business can buy and what you can trust.

What is not known yet

• Whether OpenAI or Anthropic has agreed to any cap.
• What the limit would be, or who would enforce it.
• What the other two scenarios from Altman include.
• Whether any government would get involved.

For now, this is a debate about ideas, not a rule. The full article was not available, so details beyond the summary are unconfirmed.