The Control Grid

08/31/2026

Chatbots learned by reading the internet. Robots cannot learn to move that way.
They need recordings of the real world. Somebody will own those recordings — companies, cities, or states.
Today’s Control Grid investigation is about that quieter fight: who owns the physical world AI learns from.

LINK:

China and the U.S. are building rival systems for embodied AI training data, raising new questions about ownership, access, workers and who controls the physical world machines learn from.

08/31/2026

This week’s Control Grid reporting followed a common thread: AI capability can scale quickly. Infrastructure, capital and legitimacy move on very different clocks.

08/28/2026

The AI buildout is becoming a test of institutions as much as technology.

Communities, utilities, companies and governments are negotiating who pays, who benefits and who gets to decid

08/26/2026

The AI boom is colliding with a problem that money and compute cannot automatically solve: public legitimacy.
The benefits of AI infrastructure are often framed nationally. The costs arrive locally through power demand, land use, water, rates and development decisions.

LINK:

08/24/2026

OpenAI can slow frontier development because of a safety concern. But the data centers, compute commitments,financing and power infrastructure being built around continued AI growth operate on much longer timelines.

Our latest investigation follows what happens when the safety system slows while the economic machinery keeps moving.

OpenAI's Safety system slowed frontier development. Hundreds of billions in compute commitments kept moving.

08/23/2026

One theme connected almost every major AI story this week:

The model is only the visible layer.

Underneath it sit semiconductors, memory, energy, financing, infrastructure, regulation and geopolitics.

Understanding AI increasingly means understanding who controls those layers

08/22/2026

The AI race isn't simply America versus China on a benchmark leaderboard.

It's increasingly a competition between technology ecosystems.

Models.
Chips.
Clouds.
Infrastructure.
Standards.
Distribution.

The countries in the middle may ultimately decide which stack travels farther.

China’s open models are changing where power lives in AI

08/21/2026

AI is moving deeper into national security.

That means the important questions aren't limited to which models are most capable.

Who controls them?
Who authorizes their use?
What systems are they connected to?
And where does accountability sit when AI becomes operational infrastructure?

The Pentagon is accelerating military AI while wrapping the strategy in frontier mythology and tactical branding.

08/21/2026

The United States has relied on private cybersecurity companies for years.

But offensive cyber operations raise a different set of questions.

As private companies become more deeply involved in the infrastructure supporting U.S. offensive cyber capabilities, the boundaries between government authority and private-sector operations become harder to define.

Who authorizes an operation?

Who controls the infrastructure?

Who carries responsibility when something goes wrong?

Today's Control Grid investigation follows that changing relationship.

Washington kept control of the targets. The unanswered question is who owns the consequences when private companies join cyber operations.

08/21/2026

AI safety systems are supposed to tell us when increasingly capable models cross dangerous boundaries.

But what happens when the evaluation system itself fails?

That's not merely a technical problem.

It's a governance problem.

Three AI companies. Three breaches. Fifteen days. The evidence points to a much bigger story.