A recent discussion around Anthropic raised a question that is becoming harder to ignore in the AI era.
If AI systems become powerful enough to help build their own successors, who decides when the world should slow down?
That question gained attention after Anthropic Institute published a piece titled When AI Builds Itself on 4 June 2026. Some reactions on social media framed it as if Anthropic were calling to stop AI altogether. But that is not the real point.
The deeper issue is not whether AI should be stopped today. It is whether the world has any credible mechanism for responding if future systems move beyond meaningful human oversight.
That matters because Anthropic is not speaking from the sidelines. The company is one of the central players in the AI race, and public reporting on the piece highlighted a striking internal statistic: more than 80% of the code being added to Anthropic’s own codebase is now written by Claude, up from low single digits around a year and a half earlier.
In other words, the phrase AI is building itself is no longer just rhetorical. It is becoming operational.
The real question is not whether AI should stop
The more useful question may be this:
If a point ever comes where slowing down becomes necessary, who gets to decide?
Would that power sit with technology companies? Governments? Regulators such as the European Union? Or a broader public process involving the people most affected by these systems?
At the moment, the answer is unclear.
And that uncertainty is exactly what makes this conversation important.
A new governance problem
For years, much of the AI conversation focused on capability: better models, faster performance, stronger reasoning and more useful applications.
Now the governance problem is becoming harder to avoid.
As AI systems become more capable, the challenge is no longer only what they can do. It is how societies would respond if the pace, autonomy or strategic impact of those systems began to exceed existing oversight structures.
That is not only a technical issue.
It is a political, institutional and global coordination issue.
Why this matters
At AI Dubliners, we often focus on where AI creates practical value across companies, sectors and communities.
But discussions like this remind us that AI is not only a product story or an innovation story. It is also a systems story.
The real issue may not be how far AI can be pushed.
It may be whether we are prepared, right now, to decide when slowing down is necessary and who would have the legitimacy to make that call.
That may become one of the defining questions of the next AI era.
Source
Anthropic warns about fully recursive self-improvement in AI
