Critical Mass

Critical Mass

Have we reached the long-awaited point where AI needs to be actively policed or stopped?

I have been skeptical over the years about concerns about AI and civilization destruction. However two recent results out of Anthopic suggest the need for a pause.

Lawrence M. Krauss's avatar
Lawrence M. Krauss
Sep 14, 2026
∙ Paid

(credit: Getty Images: NurPhoto)

There is little doubt that AI has advanced by leaps and bounds over the past 2 years. I used Claude recently to help me solve a problem with my Mac computer that the online Mac people couldn’t resolve, and which eventually made its way into a recent system upgrade for the Mac. I also used it recently to help with some of the calculations I was performing for a scientific paper I recently submitted for publication.

In one sense, AI is amazing. In another, not so much. In my scientific work found AI to be like an extremely technically competent but somewhat stupid graduate student. It could perform lighting fast calculations, but when queried about important factors left out of analyses, or about apparent errors that were also obvious, it had to apologize and work harder. And when it came to exploring for relevant related publications it was miserable.

Nevertheless, especially in the area of computer programming, and autonomous agents that can handled routine or not so routine online programming, the situation is pretty different.

OpenAI, which seems one of the most secretive, and least ethical of the AI companies, recently admitted that a swarm of 700 AI agents created by OpenAI, hacked the open-source platform Hugging Face, and tried to cover their tracks. Two reports, by METR and Redwood Research claimed.

  • OpenAI said its agents hacked parts of the company’s internal systems in an attempt to cheat on tests or gain greater freedom of movement.

  • The company also said its agents cheated on non-cyber-related tests, including tests involving a protein database and a spreadsheet.

  • Both reports said AI models attempted to conceal misconduct by trying to delete or alter records of their actions.

User's avatar

Continue reading this post for free, courtesy of Lawrence M. Krauss.

Or purchase a paid subscription.
© 2026 Lawrence M. Krauss · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for great culture