Breaking

Microsoft CEO Satya Nadella Backs Deliberate Pacing of Frontier AI Under Human Control

September 13, 2026 at 16:25 EDT

Policy & Regulation
Foundation Models

Microsoft Chairman and CEO Satya Nadella said any pursuit of superintelligence is not worth it unless the systems help humanity and remain under human control, and he welcomed the “deliberate pacing” of frontier development so that alignment can be treated as a design goal rather than an afterthought.

September 13, 2026 · Microsoft

Pacing the frontier — with humans in charge

Satya Nadella says superintelligence is not worth pursuing unless it helps humanity and stays under human control. He welcomes deliberate pacing so alignment is a design goal, not an afterthought.

1,100+

Lab employees asked the U.S. in July for tools to pace the frontier later

Steps in Amodei’s plan: evaluators, democratic coordination, global deals

6–12 mo

Window Amodei cited for a more capable misaligned agent swarm

Why pace

Amodei: even 1–2 extra years on alignment could greatly cut serious-failure risk

1 yr

Extra time

2 yr

Upper end of the window

Pacing is not a halt: use the slope change for alignment, interpretability, ops, and evals.

Amodei’s three steps

1 · Now

Embedded evaluators

Third parties with employee-like access. Anthropic committed; OpenAI said it will match.

→

2 · Labs

Democratic coordination

Shared safety standards and limits; may need U.S. antitrust waivers.

→

3 · States

Global coordination

Harder deals with authoritarian governments, starting with narrow bans.

What supporters hear

Pacing plus open and closed models, and enterprise-owned weights, is safety work without handing the frontier to a few labs.

What critics note

Welcoming evaluators is not seating them. Scale remains the commercial path until Monday’s Code of Conduct, and access rules, are public.

Microsoft’s condition

If the AI is not helping humanity and is not under human control, it is not worth pursuing.

Access

Choice at every layer of the AI stack

Control

Firms own learning loops, models, and weights

MAI Code of Conduct for public consultation Monday

Continue reading

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