November 4–5 — Democrats in the U.S. Congress are pushing draft legislation, the “Preventing an Artificial Superintelligence Act,” that would prohibit any individual or organization from developing or deploying artificial superintelligence (ASI), per reports from Fast Technology, Sina Finance and others. Penalties are severe: implicated companies would be forcibly dissolved, and individuals face up to 20 years in federal prison — sentencing calibrated to that for illegal nuclear-weapons development.
The signal followed the midterm elections. On November 4, Arizona Representative Yasmin Ansari said Democrats should put “pausing the development of advanced AI” at the top of the agenda. With Democrats beating expectations in the midterms, as Cailianshe and others noted, AI oversight is climbing the political list alongside the results.
Key provisions (draft)
Prohibited: developing or deploying artificial superintelligence (ASI) by any individual or entity
Individual penalty: up to 20 years in federal prison
Corporate penalty: forced dissolution of the implicated company
Sentencing benchmark: illegal nuclear-weapons development
By design this is prohibition, not guardrails — no licenses, no graduated tiers; superintelligence development becomes a criminal target outright. Current U.S. debates center on deployment safety, export controls and state-level compliance such as California’s SB series; framing “development as crime” in federal legislation is a first for a party-level push.
The legislative road remains long. The bill must clear both chambers and the president’s desk, and any step can rewrite or shelve it; even within the Democratic caucus, a blanket “pause advanced AI” stance is far from settled. Ansari’s post is an agenda call, not a caucus decision.
Markets moved first. The Nasdaq fell more than 4% on November 5 with AI names sold off, as investors weighed midterm outcomes and regulatory risk on the same page — stacking on top of capex expectations to amplify volatility.
For China’s industry the draft binds little directly, but the signal is clear: the safety boundary around superintelligence is entering legislative language. If a “development ban” becomes one side’s formal position, global AI governance gains a reference frame anchored at prohibition rather than a contest of efficiency and scale.
The bill’s formal text, co-sponsor list and committee schedule are the next markers of real momentum.