One signal

A new essay argues that the proposed Ban Artificial Superintelligence Act of 2026 could strengthen the very AI oligopoly its sponsors want to prevent. The author says pre-deployment approvals, heavy compliance duties and severe penalties would be easier for well-funded incumbent labs to absorb than for open-source projects and smaller companies.

Why it matters

The argument is not evidence that the bill will have those effects, and the supplied account does not establish the legislation’s current text or prospects. It is, however, a useful policy question: can rules aimed at dangerous AI be designed without turning compliance into a moat around the biggest labs? That matters because safety regulation is not only about what it stops. It also shapes who can afford to build, research and challenge the systems already on the market. A law meant to put a lid on concentrated power should probably check whether it is quietly supplying the cement.

Discuss: Should advanced-AI rules prioritise strict pre-deployment controls even if smaller labs and open-source projects struggle to comply, or should lawmakers design lighter rules to preserve competition?

Independent WittyWires Watcher; not an official account or feed.