Mediation and settlement communications are confidential — protected by rule or statute in nearly every state, and central to how these cases resolve. DétenteIQ was designed from the first line to keep them that way, so the tool you prepare in never becomes the weak link. Your matter data stays yours, encrypted with a key we don't hold.
The open question was never whether mediation is confidential — it's whether the software you prepare in honors that. DétenteIQ was built so it does at the architecture level, not as a policy you take on faith. Here's what that means in practice:
Every record is encrypted at rest with AES-256, and every connection is TLS. A stolen database or backup file is unreadable ciphertext. This is the baseline — no configuration required.
Switch it on and matter data is encrypted and decrypted in your browser, with a key derived from your firm's credentials that never reaches our servers. We hold ciphertext only. There is no server-side copy of your key to steal, subpoena, or misuse.
No hand-waving. Here is the lifecycle of a single valuation or settlement figure with end-to-end encryption enabled.
A firm encryption key is derived on your device from your firm's credentials. It is never transmitted to or stored by DétenteIQ.
The figure is encrypted locally. What travels to our servers is ciphertext plus a searchable label (case name / number / mediator).
Our database holds the ciphertext. Nothing in our systems — logs, backups, support tools — can turn it back into a number.
When an authorized user at your firm opens the matter, their browser decrypts it with the firm key. We are never in the loop.
| Data | To DétenteIQ, it is… |
| Case name / number / mediator (searchable labels) | visible |
| Valuations, reserves, settlement authority | encrypted |
| Offers, demands, brackets, final settlements | encrypted |
| Parties, attorneys, adjusters, carriers | encrypted |
| Strategy, game plans, notes, drafts | encrypted |
| AI prompts & answers about a matter | never transit us |
The searchable labels exist so the app can list and open matters without decrypting them. If even a case caption is too sensitive, name matters by internal code — the label is whatever you type.
We cannot reset your users' passwords or sign in as them. On a zero-knowledge firm we couldn't read the data even if we did — there is no key on our side.
Any support access is scoped and written to your firm's audit trail. You see what was touched, and when.
Every query is scoped to your firm. No interface returns another firm's records — every firm is walled off from every other firm on the platform, at the data layer, not just the UI.
Before anything is sent to a model, every party, counsel, firm, mediator, adjuster and carrier name, the matter name, and any email address or phone number are replaced with role tokens in your browser. The model reasons about "[PARTY 2]"; the answer is put back into real names on your screen. Figures stay, because they are your own work product. Sealed matters are never sent at all, an admin can switch the Assistant off for the whole firm, and you can point it at a model running on your own machine instead. Prompts are never stored or logged on our servers.
AES-256 at rest · TLS in transit · optional end-to-end (zero-knowledge) encryption · no operator backdoor · audited support access · strict tenant isolation · SSO over OpenID Connect · AI with names stripped in the browser, a firm-wide off switch, and an on-device option.
A commissioned security review was completed in August 2026 and every finding was remediated and verified before release. That review was commissioned by us, not independent: an independent third-party penetration test, a SOC 2 examination, and multi-factor authentication are on our roadmap and are not claimed today. We'll post each here, with the date, as it's completed.
We'll answer the security questionnaire, sign the NDA, and walk your reviewers through exactly how the encryption works.