01
Heppner v. SDNY, 2026.
A federal court ruled consumer AI prompts are not privileged. Discovery now reaches your AI history.
Design-Partner Beta is open
Private processing. Protected privilege.
Safely use AI with sensitive documents and live video meetings with verifiable audit logs and opt-in access to frontier cloud models.
v1.0 · Beta — open · macOS first
Trusted by teams navigating
- VA Legal
- Federal Public Defenders
- Am Law firms post-Heppner
- Private Credit Funds
01 — Problem
What changed.
A short read of the regulatory, professional, and insurance landscape now shaping how sensitive work gets done.
02
30+ state bars now require AI supervision.
The ABA and most state bars added auditability and supervision rules in the last 18 months.
03
LPL carriers added AI exclusions.
Renewal questionnaires now ask for a written AI policy. Some carriers exclude AI-related claims outright.
04
Part 2 protections do not survive the cloud.
Therapy notes and SUD records lose heightened protection the moment they enter cloud AI. (See Illinois AI Therapy Act.)
05
One paste creates MNPI exposure.
Term sheets and borrower data pasted into consumer AI surface in LP diligence and SEC exams.
06
Consumer AI is already banned in public sector.
The VA, federal defenders, and most courts prohibit consumer AI tools for case work.
Most teams are working around this with shadow AI. Sotto removes the need to.
02 — Workspace
How sotto works.
Four functions working together: Redact, Route, Convert, Audit. Local AI by default. Cloud models when you choose. Always auditable.
The primitives are open source. Sotto packages them into a compliance-ready workspace. See the architecture ↓
01
Redact
Automatically detects and protects sensitive content (privilege, PHI, NPI, MNPI) before it ever leaves the machine.
02
Route
Decides whether a request stays fully local or, with your explicit approval, calls a frontier cloud model.
03
Convert
Generates a high-fidelity synthetic replica of sensitive content so you can safely use frontier cloud models without exposing real data.
04
Audit
Cryptographically logs every action (local or cloud) so the full chain of work stays verifiable in court, on exam, or in diligence.
Sotto
ogentic-audit · verify
03 — Surfaces
Documents and live meetings.
Sotto protects both surfaces where sensitive work actually happens — the files you open and the conversations you join.
Documents & text
Upload sensitive material. Work on it locally.
Privileged filings, case notes, CIMs, term sheets, borrower financials, session notes. Drag them into Sotto and draft, summarize, or analyze without anything leaving your machine.
Live meetings
Join Zoom, Google Meet, or Teams under your control.
Real-time transcription, redaction, and audit for sensitive calls: intake interviews, therapy sessions, deal diligence, depositions. Every word stays on your machine by default.
04 — Architecture & Proof
Verify, don't trust.
Sotto is built on five open-source primitives. Read the code, run the verifier, and confirm the privacy claims yourself under Apache-2.0.
shield
Shield
Policy engine. Classifies content sensitivity and enforces what may leave the machine.
View source →redact
Redact
Detects and protects privilege, PHI, NPI, and MNPI before any model call.
View source →audit
Audit
Cryptographically signed, hash-chained log of every action (local or cloud). Externally verifiable.
View source →route
Route
Decides, with explicit approval, whether a request stays local or calls a frontier cloud model.
View source →convert
Convert
Generates high-fidelity synthetic replicas so cloud models work on safe surrogates, never real data.
View source →
The industry says "trust us." Sotto says: download the CLI and verify it yourself.
$ ogentic-audit verify ./matter-2024-CV-3047.log
Run the verifier yourself →The primitives are open source under Apache-2.0. The Sotto Desktop, Sotto Notes, and managed deployment are commercial.
05 — Who it's for
Who works in sotto.
01
Solo & Small-Firm Lawyers
≤10 attorneys
Trial lawyers, boutique litigation, IP, and family-law shops handling privileged work.
When this matters
Post-Heppner client alerts, NY audit rule, LPL questionnaires.
"You've answered an LPL renewal asking about AI in the last 90 days."
02
High-Sensitivity Therapists
& Behavioral Health Practices
Clinicians and group practices with PHI, SUD, and Part 2 records.
When this matters
Part 2 notes, Illinois / Nevada / Utah AI therapy laws, HIPAA documentation.
"Local execution is the only way to preserve heightened protections."
03
Private Credit & Direct Lending
Funds, BDCs, ad hoc lender groups
Analysts and PMs handling term sheets, borrower financials, and sensitive deal data.
When this matters
MNPI handling, LP diligence on AI usage, SEC focus on data flow.
"Term sheets and borrower data belong on the analyst's machine, not in a chat window."
04
Public Sector & Institutional Legal
VA, Federal Defenders, Courts, Bars
Government legal teams under explicit consumer-AI prohibitions.
When this matters
Explicit consumer-AI bans + need for court-defensible records.
"Your CIO blocked consumer AI but your team still needs AI to do the work."
Built for teams where one wrong paste is a regulatory event. Not for general consumer use.
06 — Difference
Why this, not that.
| Capability | Cloud-first AI | DIY local tools | Sotto |
|---|---|---|---|
| Model Quality | Frontier | Limited | Frontier (local 70B + opt-in cloud) |
| Data Leaves Your Machine | Always | Never | Only when you explicitly approve |
| Live Meetings (Zoom / Meet / Teams) | Cloud transcription | — | ✓. Sotto Notes, local with audit |
| Externally Verifiable Audit | —(vendor-attested) | — | ✓open-source CLI |
| Open-Source Primitives | —(black box) | Varies | ✓. Shield, Redact, Audit, Route, Convert |
| Privilege / Part 2 / NPI Safe | — | Partial | ✓by architecture |
| Vertical Specialization | Some | — | Legal, Behavioral Health, Finance |
| Regulatory Fit | Cloud-bound | DIY | Designed for it |
Every other regulated AI tool is cloud-first. None offers fully local execution + cryptographically verifiable audit. That is genuine white space, especially for private credit teams managing MNPI handling.
07 — Pricing
What it costs.
Sotto Notes
AI notetaker for Zoom, Meet, and Teams. Local transcription, no raw recording in the cloud.
$99
/mo per seat
- Joins Zoom / Meet / Teams
- Real-time local transcription + redaction
- Privileged-safe summaries
- Same audit log as Sotto Desktop
Sotto Desktop
Local AI by default. Cloud models when you choose. Always auditable.
$299
/mo per seat
Beta now
- Local Llama 3.1 70B
- Convert: safe cloud routing via synthetic replicas
- Verifiable audit log
- Vault + secure wipe
Sotto Server
Multi-user firms, shared vault + team audit
$999
/mo
Coming soon
- Shared team vault
- Centralized audit
- Role-based supervision
- On-prem deployment
Sotto Appliance
Courts, public defenders, government legal
Custom pricing
Institutional pilots
- Air-gapped hardware
- FIPS-validated crypto
- Dedicated support
- Custom integrations
Design partners get locked-in beta pricing for 24 months + direct input on roadmap.
08 — Library
The library. One form.
Twenty-one white papers covering regulatory requirements, competitive positioning, buyer checklists, and the architecture of verifiable AI.
Remove your regulatory risk.
Join the limited Design-Partner Beta. 15–20 organizations only. 90-day test against your real documents and live meetings, whether you're a lawyer, therapist, or private credit analyst.
Spots are limited. We prioritize teams with a clear regulatory, insurance, or LP audit pain point.