Breaking

Researchers Propose Delaying Arrival of Superintelligence Until 2040

July 9, 2026 at 05:07 EDT

Policy & Regulation
Foundation Models
Research & Papers

Six researchers at the AI Futures Project on July 9, 2026 published a new essay arguing that the arrival of superintelligence should be deliberately delayed until 2040. The proposal rests on slowing research, boosting transparency at AI labs, distributing power, and reaching an internationally verified slowdown centered on the United States and China.

Jul 9, 2026 · AI Futures Project

Hit the Brakes: Delay Superintelligence to 2040

Six researchers argue superintelligence should be deliberately postponed — buying society a buffer of "many years, maybe a decade or so" to prepare, verify, and solve safety problems before it arrives.

2040

Proposed target date to delay superintelligence

~10 yr

Buffer sought against rapid-progress scenarios

Pillars: slow, transparency, power-sharing, verified pact

The technical alarm behind the proposal

Anthropic flagged that recursive self-improvement — AI building its successors — is becoming more real. Two internal figures:

Code merged into production written by Claude

80%+

of production
code merged

Engineer code-shipping volume vs 2024

1×

2024

8×

now

Takeaway: shipping is up roughly eightfold — Anthropic calls for a "brake pedal" option.

A shifting call — from "pause" to "curb superintelligence"

Mar 2023

Future of Life Institute — Open letter for a 6-month+ halt to training beyond GPT-4 (33,000+ signatures)

Oct 2025

Future of Life Institute — Ban superintelligence development until scientific & public consensus

Jun 2026

Anthropic — Flags self-improvement progress and the need for a "brake pedal"

Jul 2026

AI Futures Project — Proposal to delay superintelligence until 2040

Support

Praised for specificity and rigor; a slowdown would give society time to prepare and solve problems before superintelligence arrives.

Criticism

Timeline seen as too short; alarm may accelerate the race; slowing risks "ceding the lead to China" and stifling innovation. Some say self-improvement is "still under human control — no need to panic."

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.