A product that reads trust-sensitive conversations should be able to answer these questions without a lawyer in the room.
What happens to a public Resistance Read
Your signal is read, not kept as yours. An approved result is held server-side for 24 hours — unlinked to any identity — solely so you can save it to an account if you choose; then it expires. We keep narrow, anonymous operational counts — which category and room were used, whether the read succeeded, and how long it took — so we can keep the engine healthy. Nothing enters a permanent account library, contact record, or marketing system without your explicit action and applicable consent. The one exception: if you choose "Email me this Resistance Read" and give explicit consent, we store that read and your email in order to send it.
What happens to a Katya™ conversation
Katya does not create a saved customer conversation or add a chat to your account, and no Katya answer is turned into a stored record on its own. If you want to keep the thinking, you take an explicit action: “Continue this as a Resistance Read” starts a new governed Read on the same situation, in the same room, and that Read is only persisted if you then choose to save it to an account. The conversation is transient either way. A limited private evaluation record may be retained to verify that a released answer passed Kataclyzim’s governed checks and to support Katya’s certification evidence. Your question is not used for advertising or model training, conversation text never enters analytics, there is no transcript in a link, and nothing is saved in your browser beyond a session turn count. Katya answers up to five substantive turns per session and five per day.
Boundaries and distress come before persuasion
If a signal contains "never contact me again," the answer is to stop — the read says so, and it will not produce a follow-up sequence for someone who asked to be left alone. Signals involving grief, crisis, or serious vulnerability get a non-sales response and, where relevant, a recommendation to hand the situation to a human. A stop sign is intelligence, not a failure.
Grounded means grounded
A read is built from what you actually provided. When a signal is too thin to interpret honestly, you get one clarifying question instead of a confident guess. The room you choose shapes language and stakes — it never invents facts, executives, budgets, or history you didn't supply.
How your input is processed
Generation happens server-side. Your input is transmitted to an AI provider to produce the read, and results must pass our quality and safety checks before they're shown — a result that fails is not shown at all. Don't paste information you aren't authorized to share.
Human judgment is part of the product
Reads are decision support for a person who knows the account — not scripts to run unreviewed, and not legal, medical, financial, or compliance advice. In regulated rooms, reads carry cautions and expect a human to apply them.
What we don't claim
Kataclyzim is compliance-aware, not compliance-certified. We do not claim HIPAA, SOC 2, GDPR, or any other certification we have not earned. Abuse protection, rate limits, and human review of anything paid are real; badges we haven't earned are not on this page.
Nothing in the product keeps a record of your work unless you ask it to. These are the only two states.
Transient processing
Signed out, nothing kept
A Katya conversation, and a free Resistance Read you do not save, are processed to produce the answer and then not retained as your records. No account, no transcript in a link, nothing in your browser beyond a session turn count.
Optional account persistence
You asked for it, so it stays
When you choose to save a Read, it becomes a record in your account with its room and date, reachable from your workspace, usable for practice and for a continuing Read later. You can archive it. What you see saved is the same outward projection you were shown.
The formal commitments are in the Privacy Policy; this page is the plain-language version of the same behaviour.
Expert provenance, scoped qualification, clear permissions, and evidence you can inspect.
◆ Provenance
Every governed output names its origin.
Product surfaces read “Expert-derived · Resistance Intelligence” with the standard version behind it. The Expert Value Framework keeps authorship, permission and reuse lineage attached. See provenance →
Certification records
Status, version and scope are readable.
Katya’s public record is read live from the certification record, never a static badge. Enterprise records carry scope, critical failures, remediation and renewal. Katya’s record →
Record integrity
What was disclosed, confirmed and changed is preserved.
In consumer enrollment, consumer-confirmed facts carry into the application once; corrections stay attributable. Private standards, weights, anchors and customer records are never published. How enrollment preserves the facts →
Workspaces, practice, the Memory Bank and proposed improvements all operate inside your organization's account. These are the controls a buyer should expect.
Account separation
Saved Reads, practice and evidence belong to your account. No other customer can see them.
Permissioned scope
People work inside the workspaces and roles your organization permits. Sensitive situations are handled within that scope, not broadcast across it.
Accountable oversight
Proposed improvements to the standard are reviewed by named experts and released as versions. Nothing your people or agents are certified against changes silently.
Auditable qualification
Certification records name the standard, version, scope and date. Katya's public record is verifiable at /certification/katya.
These ground rules govern every Kataclyzim certification engagement and apply to humans and AI agents alike.
Certification is evidence-based — and can be refused
Certification means the subject demonstrated the required judgment against governed scenarios and evaluation standards — meeting a defined Kataclyzim or enterprise-governed standard at the time of evaluation. Training is not proof; passing the evaluation is. NOT CERTIFIED is a legitimate outcome: if the evidence does not support readiness, Kataclyzim does not certify. Nothing is auto-granted for completing training.
Certification is point-in-time — revalidation keeps it honest
A certification reflects performance at evaluation time. People change roles, models drift, standards evolve — so monitoring and revalidation are part of the design, and a certification can lapse or be revoked when the standard is no longer met. No certification guarantees future behavior.
Context and safety behavior are part of the standard
Judgment is evaluated in the environment where it will be used — the same words can require different interpretation and boundaries by room. Knowing when not to act is part of readiness: boundary, distress, and stop behavior are scored, and human oversight remains part of every consequential workflow.
Standards stay attributable to their experts
Every governed standard has an identifiable human origin, and expert-authored standards remain attributable to the people who created them. Provenance is preserved through review, versioning, and publication.
Not a government credential
Kataclyzim certification is not government accreditation, professional licensure, regulatory approval, or legal authorization — and we won't imply otherwise unless such status is actually obtained. It is a governed, evidence-based readiness standard, and it's presented as exactly that.
The formal versions: Privacy Policy · Terms of Use · Security · Acceptable Use. Questions: use the contact form.