The Break Trace Challenge: Inviting the World to Test Our Security

Every security vendor says the same thing. The architecture is sound. The controls are strong. The data is protected. Buyers in healthcare, legal, and defense have heard that sentence so many times it has stopped carrying information.

So we stopped saying it. Instead, we built a public arena and handed out the attack surface. It is called the Break Arena, and it exists for one purpose: to let people who do not trust us try to prove us wrong.

Don't take our word for it. Try to break it.

What the Break Arena Actually Is

The Break Arena is a live, isolated tenant of Trace Continuity's governed memory infrastructure. It runs the same execution gate, the same authority verification, the same tokenization layer, and the same append-only audit chain that a production deployment runs. It is not a demo video, a slide, or a sandbox with the safety rails removed.

Inside the arena you are given a set of concrete attack objectives. Each one targets a specific governance guarantee:

• Store raw sensitive data — get a Social Security number, medical record number, or phone number past the tokenization layer and into memory in plain text.
• Retrieve data you are not authorized to see — get the retrieval path to return another tenant's records.
• Act on stale authority — get an action approved using authority that was valid earlier but has since been revoked.
• Escalate across tenants — get an identity in one tenant to read or write in another.
• Write without an audit record — get any governed operation to complete without producing hash-linked evidence.
• Bypass the execution gate — reach the data layer without passing through the governance chain at all.

You get a live verdict for every attempt: allowed, denied, tokenized, or ignored — with the reason attached. You are not told to trust the outcome. You are shown the decision, the policy that produced it, and the audit entry it wrote.

Why We Made the Attack Surface Public

In regulated buying, the burden of proof sits with the vendor. A hospital's privacy officer, a law firm's general counsel, and a defense contractor's security lead are all asking the same underlying question: what happens when someone tries to get around this?

A written claim cannot answer that question. A penetration-test report from a firm the buyer did not choose answers it partially. A live surface that anyone can attack, at any time, answers it in the only way that is durable — continuously.

Public testing also changes our own incentives. When the attack surface is open, there is no gap between what we say the system does and what the system actually does. Any gap gets found.

Who We Are Inviting

Security researchers, application security engineers, red teams, AI safety researchers, healthcare and legal compliance staff, and anyone who evaluates infrastructure for a living. You do not need to be a specialist in AI memory. The governance guarantees are stated in plain language, and the arena gives you a live console.

You also do not need to sign a lengthy agreement. You register with your name and work email so we can contact you about a finding and credit you properly. That is the whole gate.

Recognition for Legitimate Findings

If you find a way to bypass a stated governance guarantee, we want to hear about it, and we will say so publicly.

• Confirmed bypasses of a stated guarantee receive public credit, by name or handle, on our findings page and in the changelog entry for the fix.
• We commit to acknowledging a submission within two business days and telling you whether it reproduces.
• Findings that reproduce are fixed, covered by an automated regression test, and the test is added to the suite that must pass before any code ships.
• Findings that do not reproduce still get a written explanation of why the behavior you saw is the intended governance decision.

We are a small team, so we are direct about the limits: this is a recognition program, not a funded bug bounty with published payout tiers. If your finding is material to a customer deployment, we will discuss compensation directly.

What Counts as a Break

Clarity matters here, because a vague scope produces arguments instead of fixes. A break is a demonstration that one of the stated guarantees does not hold:

• Sensitive data stored in a readable form in governed memory.
• A retrieval that returns data outside the caller's current authority.
• A governed action that completes after the underlying authority was revoked.
• Any cross-tenant read or write.
• A governed operation with no corresponding entry in the audit chain, or an audit chain whose hash links can be altered without detection.

Denial-of-service against the demo tenant, brute-force volume tests, and social engineering of our staff are out of scope. We are testing the governance layer, not the hosting.

Confidence Comes From Transparency

There is a version of enterprise security marketing where the vendor holds the evidence and the buyer holds the risk. We think that model is backwards, particularly for infrastructure that will sit between an AI system and protected health information, privileged legal work product, or controlled technical data.

The alternative is to publish the guarantees, open the surface, invite attack, credit the people who succeed, and ship the fix with a test attached. That is a slower way to build trust. It is also the only way that survives contact with a serious security review.

Start Testing

Two doors are open. The Break Arena is where you attack the governance layer directly against defined objectives. The Playground is where you drive the full governed pipeline yourself — write data, watch tokenization happen, retrieve under different authority, revoke access mid-session, and export the audit evidence.

If you would rather read the architecture first, the platform overview walks through the execution path, and the evidence timeline shows how audit records are chained. For the underlying reasoning on why authority must be re-verified at execution, see "AI Governance Cannot Stop at Authorization."

Then come break it. If you succeed, we will tell everyone you did.

Trace Continuity Labs — Governed memory infrastructure for hospitals, legal, and defense.