How We Build

Evoked builds governance architecture for AI. We also use AI to build it.

What We Use

We use Claude by Anthropic. It is our AI infrastructure - development, analysis, agent operations. We are paying customers of Anthropic. We run on commercial infrastructure - Vercel, Railway, standard tools.

We are customers of the AI industry we critique.

What We Don't Do

• We don't build foundation models
• We don't train on user data
• We don't employ data annotation labor
• We don't operate data centers

We don't extract. That's the line.

What We Build Instead

Governance frameworks that any AI system can be held to. Sovereignty assessments that measure whether technology honors its users. Restraint specifications that define what agents must refuse. The architecture that tells the technology where to stop.

The Position

The technology has utility. The political economy of its production is where the harm lives. We're trying to change the economy, not pretend the technology doesn't exist.

We evoke. We never extract. That applies to how we build, not just what we build.

See our Defensive Patent Pledge.