About Advanced AI Society

Our purpose

A world where people can trust AI.

Our mission

To make open verification the default for high-risk AI systems.

How

We unite market leaders and deployers with insurers, policymakers, and civil society to standardize, deploy, and scale open verification, an AI security approach purpose-built for the agentic age.

Advanced AI Society is the alliance for verifiable AI. We unite market leaders and deployers with insurers, policymakers, and civil society to standardize, deploy, and scale open verification, an AI security approach purpose-built for the agentic age.

We work through two pillars.

Pillar 01

Open Verification Lab →

The Open Verification Lab builds research, standard, and tools to make AI safety possible in deployment.

Pillar 02

Ecosystem Growth →

Ecosystem Growth builds the category of verifiable AI.

Founding Beliefs

The Machine-to-Machine Era Demands Machine-Speed Trust.
The line has already been crossed. Autonomous agents interact primarily with other machines at speeds human oversight cannot match. Traditional governance, manual reviews, and quarterly audits operate above the surface while operational risk accumulates out of sight below. All periodic audits fall behind and every operator’s log remains a claim. The answer is not more human oversight. It is automated mechanisms carrying human authority at machine speed, emitting tamper-evident evidence anyone can verify after the fact, at will and without permission. Learn more on Why Now →
Verification Cannot Belong to the Verified.
Right now, the AI industry has expanded in the right direction from claims-based AI to independent evaluators conducting the verification. But verification cannot stop there because verification cannot belong to the verified. Trust requires open records anyone can verify without operator permission, backed by a third-party ecosystem that certifies the machinery producing them.
Open Verification Is the Architecture for Verifiable AI.
Open verification is the next expansion for high-risk AI that impacts our lives and livelihood. Open verification extends the open-source premise to the agentic stack: you cannot secure what you cannot inspect, and you cannot delegate what you cannot openly verify. Because open verification evidence is binary and deterministic, it travels seamlessly across deployers, insurers, and regulators. Every institution that we rely on to settle what happened, from courts and journalism to cybersecurity incident response, depends on evidence somebody outside the story can openly verify.
The Technology Is Here; Now We Must Make It Legible as Shared Public Infrastructure.
We need to build the open verification ecosystem for the machine-to-machine world. The technical primitives of machine trust are already built and shipping. What is missing is the architecture turning scattered verifiable AI solutions into shared public infrastructure, a single open standard that policy can mandate, procurement can require, accredited assessors can certify against, and civil society can trust. And the party doing that certifying is a third party: not the supplier, not the buyer, and never one who vouches for the output.
Building a Third-Party Open Verification Ecosystem Prevents Loss of Control.
AI could help humanity take on problems we cannot solve on human timelines alone. The same capability cuts both ways: unverified agents acting at machine speed are a civilizational risk, compounding errors faster than anyone can find them. Building a third-party open verification ecosystem is the pragmatic engineering solution that lets enterprises, governments, and communities delegate real-world authority to AI while prioritizing AI safety and system security.
Securing Machine Trust Is a Social Movement, Not Just a Technical Specification.
Ensuring we can trust AI systems that make unpredictable decisions at speeds no human can follow is a civilizational challenge that cannot remain trapped in technical whitepapers. Engineering open standards is only half the work; the other half is field-building. Broad public literacy, active communication, and convening storytellers, journalists, CISOs, and advocates are how society comes to understand, demand, and enforce open verification.
Pragmatic, Pluralistic Approaches Are the Way Forward.
Achieving machine-speed trust demands a “yes, and” approach across open source, open weight, proprietary models, local nodes, and cloud deployments. Solutions must address catastrophic risk and liability while simultaneously providing the trust foundation that makes high-stakes AI adoption and innovation possible. They must be interoperable with the current infrastructure and leverage the latest innovations from cryptography.

Our Goals

These six goals are the outcome test: they tell us when we are done. When all six are met, open verification is the standard of care for high-risk AI.

the market adopts it

The field requires shared, published definitions across three core concepts: verifiable AI, the territory we steward; open verification, the category we coined; and Proof-of-Control, the open standard we author.

What success looks like: industry participants cite these exact definitions in specifications, procurement documents, and policy papers rather than inventing proprietary terminology, making it effortless to tell a verifiable system from a claims-based promise.