OpenAI's Chief Scientist Proposes: Set an "Industry Cap" on AI Development Until Understanding Can Keep Up
OpenAI Chief Scientist Jakub Pachocki proposed that AI companies should reach a consensus to set limits on developing increasingly powerful models until researchers find better ways to understand and monitor model behavior.
OpenAI Chief Scientist Jakub Pachocki has proposed that AI companies should reach a consensus to set limits on the development of increasingly powerful models until researchers find better ways to understand and monitor model behavior.
According to an article he published on Sunday, Pachocki called for the establishment of industry-wide standards that set thresholds for AI development, with companies committing not to cross those thresholds until specific safety requirements are met.
"I further propose establishing common caps based on perplexity and other measurement metrics where developers commit not to cross these caps until 'the ability to understand and monitor AI behavior' reaches a specific milestone," Pachocki wrote.
The proposed limits could be based in part on a technical metric called "perplexity" which measures how accurately an AI model predicts the next segment of text. Pachocki did not specify what the initial cap should be set at, or when it would take effect.
He warned that AI could accelerate its own R&D process as early as 2027, potentially outpacing the speed at which researchers' ability to monitor and understand the technology improves.
OpenAI CEO Sam Altman expressed support for the proposal, calling Pachocki's article "the most important release from OpenAI in some time." In a post on X, he said that agreeing on common caps would help ensure "safety and control."
The timing of the proposal is intriguing the United States is preparing for a new round of debate on AI regulation. The Democratic Party's landslide victory in the midterm elections, with wins in House, Senate, and gubernatorial races, could lead to increased regulation of tech companies in Congress and renewed legislative efforts to establish AI safety rules.
Some lawmakers are also discussing the creation of a dedicated Senate committee to coordinate AI policy. Meanwhile, President Trump has consistently opposed broad restrictions on AI development setting the stage for a confrontation between the White House and the incoming Democratic Congress over the approach to regulation.
Related Articles

After a brutal week for US stocks, valuations now offer positioning value, but no one dares to call the bottom yet.

Title context: Goldman Sachs' "fancy move" to raise financing for Musk to buy GPUs
Text:
Goldman Sachs' "fancy move" to raise financing for Musk to buy GPUs

Regulators unveil a package of measures to stabilize the market: lowering the margin ratio for margin financing, introducing market makers on the ChiNext board, and implementing multiple fee reductions simultaneously.
After a brutal week for US stocks, valuations now offer positioning value, but no one dares to call the bottom yet.

Title context: Goldman Sachs' "fancy move" to raise financing for Musk to buy GPUs
Text:
Goldman Sachs' "fancy move" to raise financing for Musk to buy GPUs

Regulators unveil a package of measures to stabilize the market: lowering the margin ratio for margin financing, introducing market makers on the ChiNext board, and implementing multiple fee reductions simultaneously.






