# GDPR Data-Flow Declaration

> GDPR data-flow declaration: every path on which data leaves IRONSTICK towards an external AI provider, what is transmitted and where pseudonymization applies.

Original page: https://www.ironstick.legal/gdpr.html

---

Relations with External Large Language Models

Factual, code-derived description of every path on which data leaves
IRONSTICK towards an external LLM, what exactly is transmitted, in which form, and
where the pseudonymisation layer applies. Windows desktop edition. This document
describes IRONSTICK's behaviour only — **what the AI provider does with received
data is governed exclusively by the provider's own terms** (the terms and
privacy policies of the providers the user has set up: Alibaba Cloud (Qwen),
Moonshot AI (Kimi), OpenAI, Anthropic and, if switched on, Groq, OpenRouter and
Mistral; Google only if Google Drive is used for exports), which the
user must review separately.

Version 1.0.7 — as of November 2026

**Summary.**

- **Storage.** All data lives in one local data directory. The database,
the document store, the configuration file and the AI scratch store are AES-256
encrypted at rest. Clear text remains only in temporary working copies of opened
documents, the scan inbox, exports and outgoing files, the short-lived
peer-review buffer and, if switched on, the diagnostic log (section 8).
- **External AI.** In-app AI calls exist only with the user's own API key.
Calls through the central router are pseudonymised by default and recorded in a
local audit log; the deliberate clear-text exceptions are listed in 2.3.
- **Consent.** The AI chat, document capture and person research ask once
per session or program run; in the word processor, starting an AI function is
the consent. Several background functions call the cloud AI without a dialog of their own:
e-mail triage, the Data Consistency Monitor, the daily briefing, the dialogue
summary and the Red/Blue Team review (section 3).
- **Desktop-AI bridge.** An optional add-on that passes real case data to the provider of the user's own desktop AI, after
one consent dialog per program run (section 4.9).
- **Other services.** Party addresses are geocoded daily with
OpenStreetMap Nominatim (address only, no names); a news file is downloaded from
[www.ironstick.legal](https://www.ironstick.legal/index.html) after start-up and every six hours; no case or personal data
is sent to the software vendor (sections 1 and 7).
- **IRONSTICK SERVER.** Optional, on the firm's
own NAS inside its local network, without internet connections. Documents travel
to the server and rest there only as AES-256-encrypted files; the server never
sees their plain text. Working time per case
and backup events of each workstation are visible to the firm's administrator
(section 9).

## 1. Scope and endpoints

![AI API, AI desktop app and MCP server on the left, the IRONSTICK Client in the middle, data vault with AES-256, IRONSTICK Server and scan app on the right](https://www.ironstick.legal/img/65_ai_client_vault.png)

In-app external AI calls go to **the provider the automatic cascade is
using at that moment**. The user can set up up to four providers; the
providers for which a key has been entered are used in this fixed order,
lowest cost first: **Qwen (Alibaba Cloud) → Kimi (Moonshot AI) →
ChatGPT (OpenAI) → Claude (Anthropic)**. At every program start IRONSTICK
checks which of them respond. If a provider fails, the same request goes to the
next one. The failed provider is retried after a waiting time that depends on
the error; it is
skipped for the rest of the day only when its credit is used up or the cause is
unknown. The content transmitted
is identical on every provider (section 3 onwards: pseudonymised by the
router, consent, audit log). The endpoints are:

- **Claude:** POST https://api.anthropic.com/v1/messages,
the user's own API key.
- **ChatGPT:** POST https://api.openai.com/v1/chat/completions,
the user's own API key.
- **Qwen:** POST https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions
(Alibaba Cloud Model Studio, international region), the user's own API key.
- **Kimi:** POST https://api.moonshot.ai/v1/chat/completions
(Moonshot AI), the user's own API key.

**Free background providers (optional).** In addition, the user can
store the free offers of **Groq**, **OpenRouter** and **Mistral**.
IRONSTICK uses them exclusively for tasks that run in the background:
e-mail triage, Data Consistency Monitor, daily briefing, dialogue summary,
speaker assignment in transcriptions, retrieval of statute texts and situation
report. Each of these providers has its own on/off switch in the
configuration; it is used only if a key is stored and its switch is on.
They are **never** used for the chat, for pleadings or for the Red/Blue Team
review. If none of them is available, the cascade described above takes over.
Pseudonymisation, consent and audit log follow the respective task
(section 4). The endpoints are:

- **Groq:** POST https://api.groq.com/openai/v1/chat/completions,
the user's own API key.
- **OpenRouter:** POST https://openrouter.ai/api/v1/chat/completions,
the user's own API key.
- **Mistral:** POST https://api.mistral.ai/v1/chat/completions,
the user's own API key.

**Free offers — notice in the configuration.** In the configuration, a notice below each of these providers points out that IRONSTICK — provided the API key is stored and the switch is on — uses it for background tasks such as summaries and e-mail triage and, in doing so, transmits data to that provider. Before switching any of
these providers on, check whether its terms provide the data-processing
agreement required for processors and meet your professional rules. A provider
that is switched off is not used.

Three configurable model tiers are used per provider (small / medium /
large). A separate, optional channel is the
**desktop-AI bridge** (section 4.9): there, data flows into the user's own
local desktop AI (Claude Desktop / Claude Code, ChatGPT Desktop, Qwen
Desktop **or** Kimi Desktop), i.e. to that AI's provider (Anthropic, OpenAI,
Alibaba Cloud/Qwen or Moonshot AI/Kimi) under the user's own subscription/account with that provider. The
bridges may run in parallel: each desktop AI has its own small bridge program,
all of them connect to the same local endpoint in the running application, and
they can work on the same case simultaneously. No case or personal data is sent
to the software vendor. The only vendor contact is a plain download of a news
file from [www.ironstick.legal](https://www.ironstick.legal/index.html) (after start-up and every six hours, section 7),
which carries no data apart from the technical connection itself.
**IRONSTICK SERVER** (section 9) does not change this: it is
operated by the law firm itself, on the firm's own hardware, inside the
firm's own network — it never contacts the vendor and never contacts any AI
provider.

**The master switch is the user's own access.** In-app
external AI processing exists only if the user has opened their own account
with an AI provider and entered their own access key in the configuration.
**Without any provider set up, no in-app external AI call can occur at all** — every
feature described in section 4 then stays local or is simply unavailable. That
account is a direct contract between the user and the provider; IRONSTICK is
not a party to it. The same applies to the desktop-AI bridge, which requires
the user's own subscription/account with the chosen desktop AI provider
(Anthropic, OpenAI, Alibaba Cloud/Qwen or Moonshot AI/Kimi). Documents marked **secret** are excluded
from the bridge's and the chat's tools, from the AI instruction
package and from every generated output (attachments, dossiers, exports);
their processing during capture follows the user's deliberate choice at
capture time (section 4.1).

Fully local (nothing leaves the machine): text extraction and OCR, the
deterministic local chat layer, global search, relevance radar, letter-chain
analysis, dossier/PDF generation, the redaction finder's local pass, speech recognition (transcription of recordings), and the building of the AI instruction package.

## 2. The pseudonymisation layer

Every in-app call through the central AI router is pseudonymised by default:
identifiers in outgoing text are replaced with tokens, and the answer is
restored locally. If the pseudonymiser cannot be built
completely, nothing is sent. The exceptions are listed in 2.3. The mapping lives
only in memory, one fresh mapping per operation; it is never stored.

![Top: IRONSTICK Client, anonymised figure with gears, AI API; bottom: IRONSTICK Client, user, AI desktop app](https://www.ironstick.legal/img/66_pseudonymization.png)

### 2.1 What is replaced

| Source | Fields | Token family (internal identifiers) |
| --- | --- | --- |
| Client records | name, national ID number from the ID field (e.g. Romanian CNP), phone, e-mail, address and city | $PARTEI_A$ (party), $CNP_A$ (ID number), $TEL_A$ (phone), $MAIL_A$ (e-mail), $ADRESSE_A$ (address) |
| Persons / entities | full name (person or company), national ID or company number from the ID field (e.g. Romanian CNP or CUI) | $PARTEI_B$ (party) / $FIRMA_A$ (company), $CNP_B$ (ID number) |
| Cases | case number, opponent, claimant | $FALL_A$ (case), $PARTEI_C$ (party)… |
| Pattern layer (unknown values in free text) | court file numbers, personal identification numbers, IBANs, e-mail addresses and phone numbers. Other national ID numbers are replaced only if they are stored on the client or person record. | same token families |

### 2.2 What deliberately stays real

- **Dates, times and monetary amounts** — they are needed unchanged for the legal analysis.
- **Institution names** (courts, authorities) — they are not in the
dictionary and are public bodies.
- **Third parties without their own record** — witnesses, doctors, lawyers, judges or companies that are not recorded as a client, person or case record; anyone who should be replaced by a token is simply recorded as a person.
- **Addresses of non-clients** — the client's stored address is the one replaced by a token.
- The entire **factual content** of documents (allegations, diagnoses,
narrative) — only identifiers are replaced, not the story.

### 2.3 Deliberate clear-name exceptions

- **Person web research (stage 1)** and **form autofill** for persons /
institutions: a web search for $PARTEI_A$ is meaningless, so the
real name is sent by design.
- **Redaction AI finder**: it receives text that has already been redacted
locally, and must answer with the clear names of *unknown third parties* it
finds — pseudonymisation is off for this one purpose.
- **Read-aloud preprocessing (TTS)**: see 4.6 — sentences go out in clear
text.
- **Desktop-AI bridge**: raw data by design (section 4.9).

## 3. Consent moments and the audit log

Calls routed through the central router are consent-gated and logged: each
model round writes a row to the local audit table with
the complete request and response **in the tokenised form that actually left the
machine**; refused or blocked attempts are logged without content. Where a
function asks for consent, the dialog states model tier and approximate request
size, and one approval covers the current screen session (for document capture:
the current program run).

**Router flows without a consent dialog of their own:**
e-mail triage (in background operation and on a manual fetch in the e-mail
screen alike, section 4.2), the Data Consistency Monitor's analysis run (4.5),
the daily briefing (4.4a), the dialogue summary of the API chat and the Red/Blue
Team review (4.7), and the redaction AI finder (enabling its toggle is the
consent). These are still pseudonymised (except the redaction finder's intended
clear-name answers) and still audited.

**Flows that bypass the router:** the word processor's
AI functions, read-aloud preprocessing and the File→Markdown AI correction
call the API directly. They are pseudonymised (exception: read-aloud),
but they show **no consent dialog of their own and write no audit-log
rows**. The user's operation of these features (pressing the AI button,
starting a generation, starting read-aloud with the AI toggle on) is the
consent moment.

## 4. The data flows, use case by use case

### 4.1 Document capture

Text extraction and OCR are local. If an API key is configured, IRONSTICK asks
once per program run, then up to four call types run (all pseudonymised, all
audited):

- **Main-document detection** (small tier): excerpts of the raw OCR text — only for poor-quality scans.
- **OCR correction** (small tier): a limited excerpt of the document text.
- **Classification** (medium tier): a limited excerpt of the document text, **plus system lists**: the complete case list (case number,
client name, opponent — tokenised), a known-names pool (tokenised), and the institution-type and document-type code lists
(untokenised labels). Fallback: small tier, without the
lists.
- **Contextualisation on save** (small tier): the document's metadata and
full text of limited length, plus an index line (ID, date, title) of **every active evidence item of the same client** — used to ground
cross-references. Runs also for documents marked secret (deliberate user
decision).

### 4.2 E-mail monitor (triage)

Case assignment is done **locally and without AI**. New mail is fetched
in the background at start-up and then every five minutes. An e-mail whose case
the local recognition already identifies (case number, letter number, client
number or reference to an own submission) causes no AI call. For every other
e-mail, **one small-tier call** assesses the urgency, with exactly: sender
name and address, subject, and the stored preview snippet as plain text
(hard-capped) — **never the full body**, no attachments
and **no case list**. Answer: priority and reasoning. If free background
providers are switched on (section 1), this call may run through them.
**No consent dialog is shown**, neither in background operation nor on a
manual fetch in the e-mail screen: the triage runs automatically once a mail
account and an AI provider are set up. Pseudonymised and audited.

**No tracking pixels.** Opening a mail in the monitor never contacts
the sender's server: images referenced by an internet address (including
invisible tracking pixels) are not loaded and appear as placeholders with a
notice, so the sender cannot learn when, how often or from which address a
mail was read. Only images embedded in the mail itself are displayed.

### 4.3 Person analysis (web research)

Three stages, after a once-per-session consent:

- **Stage 1 — search:** search engines are queried directly from the
user's computer; additionally the
provider's server-side web-search tool is invoked with the person's **real
name in clear text** (unavoidable for a search; logged in clear text).
- **Stage 2 — page summaries** (medium tier): per page, the person's name
(tokenised), the URL and a limited excerpt of the page content.
- **Stage 3 — profile synthesis** (large tier): name and city (tokenised)
plus the collected page texts, of limited length per source.

The result is stored locally with reliability "unconfirmed". The related
**autofill** buttons on the person/institution forms likewise send the typed
name in clear text (documented exception).

### 4.3a Public-source party check

The public-source check of a party is a fixed procedure the application runs
itself when a connected AI calls it with name, city, district, country and
the kind of party (natural person or legal entity):

- **Case files:** a full-text search for the name across the local
database — **purely local**, nothing leaves the machine.
- **Internet:** a fixed list of search queries is sent to the public
search engines **Google, DuckDuckGo and Qwant**. Each query consists of a
search term in the language of the party's country plus the party's **real
name in clear text** and, for some queries, the city or district
(unavoidable for a search). No case, client or document content is
transmitted. The engines see the queries like those of any browser user.

The result is returned to the calling AI as a prepared, unverified profile.
Nothing is written into the master data: every finding passes the review gate
of the Data Consistency Monitor (section 4.5) and is adopted only when the
user accepts it. Only publicly accessible sources are evaluated.

### 4.4 Daily log

The screen itself is local. Importing free-form text runs a **local**
parser first; only when no daily-log format marker is found does an AI fallback
normalise the raw text: the pasted text is sent completely, small tier, consent-gated, pseudonymised, audited. Stored
daily-log entries leave the machine only as tool results of the chat or the
desktop-AI bridge (capped excerpts) or inside the AI instruction package
(section 5).

### 4.4a Daily briefing

On the first program start of a day (Windows), IRONSTICK
prepares a spoken daily briefing in the background, already during start-up and
before the briefing window asks whether the user wants to hear it. The briefing
text — deadline counts, today's and tomorrow's court dates with case number,
client name, city and court, and the titles of today's appointments — is sent
once (small tier) to the cloud AI, which rewrites it as spoken prose in the
system language. Pseudonymised through the router, audited, **no consent
dialog**. If free background providers are switched on (section 1), this call
may run through them. Without an available provider, a locally assembled text
is used instead.

### 4.5 Data Consistency Monitor

Phase 1 (e-mail ↔ master data) is fully deterministic. Phase 2 sends
**one small-tier call per new inbound document** (pseudonymised, audited,
**no dialog**) containing: the de-duplicated **contact-region snippets**
(short excerpts around address, phone and tax details) from the full text of
**that** document, the party names of that document and of its case, and, for
each party, the **names** of possible matches from the stored master data
(tokenised where in the dictionary; institution names in clear text). **Contact
data and ID numbers from the stored master data are not transmitted.** If free background
providers are switched on (section 1), this call may run through them. The
answer only ever becomes a suggestion that the
user must accept in the monitor.

### 4.6 Word processor (AI assistance in the text)

- **Revise selection:** the selected sentence/paragraph/section (with
markup) **plus an excerpt of the document as context**
and the user's instruction — medium tier, pseudonymised.
- **Structure/TOC:** effectively the whole document as an indexed line list
(long lines truncated), of limited length — medium tier, pseudonymised.
- **Read-aloud preprocessing:** when reading a document aloud, **every
sentence is sent individually to the small tier in clear text** to expand
abbreviations and numbers — **without pseudonymisation** (identification
numbers must return unchanged to be read correctly). File stamps and reference
markers are stripped locally first. Starting read-aloud is the consent
moment.

None of the word-processor calls shows a separate consent
dialog or writes audit-log rows (router bypass, section 3).

### 4.7 Red/Blue Team review and dialogue summary (API)

Two further API flows belong to drafting. **Red/Blue Team:** if the user
has switched it on for the active provider (switch next to the API key; a saved
key is required), every formal document handed over through the delivery
dialog — from the in-app chat and from the desktop-AI bridge alike — is read
once more by the same provider as opposing counsel in a separate review pass
(medium tier). Transmitted: the draft text, the date and the case's
jurisdiction; pseudonymised through the router and audited like every router
call; **no consent dialog** — the switch is the consent; the reviewer's report is attached for the user. The Red/Blue Team review
**never** runs through the free background providers (section 1). **Dialogue summary:** in long
API-chat sessions the older part of the dialogue (only the user's questions and
the model's final answers, never tool results) is condensed by the active
provider (small tier, background, **no consent dialog of its own**) into a
short summary that replaces the raw history; it is stored encrypted
in the scratch store and overwritten each time. Pseudonymised and audited. If
free background providers are switched on (section 1), the summary may run
through them.

### 4.8 AI chat (external, API)

Consent once per chat session; medium tier with automatic escalation to the
large tier (the whole context is then sent again); pseudonymised; every round
audited. Each turn transmits: the system prompt (static rules), the **live GUI
context** (current screen, open case/client incl. client name — tokenised),
up to **10 AI-memory entries** of the open client, the most recent dialogue plus a summary of older turns
(4.7) and the list of the chat's scratch files. Every tool the model calls
returns its result into the API conversation — including document full texts
(larger results are stored in the encrypted scratch store), e-mail bodies and
daily-log excerpts. The chat uses the same tool layer and the same delivery
gates as the desktop-AI bridge (4.9): pleading verification, Red/Blue Team (4.7)
and hand-over through the export dialog — what leaves the system is a
submission-ready template. The local chat layer answers routine requests
without any transmission.

Two utility functions deserve explicit mention. The **weather lookup**
fetches current weather data for a named place from the non-LLM service
Open-Meteo (section 7) — only the place name / its coordinates are transmitted
to that service, never case or person data. The **location lookup**
answers exclusively from the **most recent locally stored** security
snapshot (section 8): the AI's request triggers **no new external lookup** —
it only reads what is already on disk. As with every tool, what these functions
return flows back into the AI conversation, i.e. to the AI provider.

### 4.9 Desktop-AI bridge (optional add-on modules)

**Each desktop-AI access is an optional module, purchased separately.**
It exists only where the practice has acquired and installed the respective
bridge module; without any module, the bridge section stays locked in the
configuration, the local endpoint never starts, and none of the flows
described in this section can occur. Bridges are available for
**Claude Desktop/Claude Code**, **ChatGPT Desktop**,
**Qwen Desktop** and **Kimi Desktop**; each has its own small bridge
program, and all of them connect to the same local endpoint in the running
application. The data flows below are identical for all; only the receiving
provider differs.

**This channel transmits real case data.** That is its purpose: the user's own desktop-AI session drafts
submission-ready templates and therefore needs real names and numbers.

- The desktop AI (Claude Desktop/Code, ChatGPT Desktop, Qwen Desktop or Kimi Desktop) connects through its bridge program to the running application; this connection stays on the computer. Only the web search, page-reading and register tools contact the internet.
- **Websites with an access check.** If a website requires a confirmation before
it releases its content, IRONSTICK opens the page in a separate window; the
user confirms there personally. The desktop AI then reads the content of the page —
it flows to its provider like any other tool result.
- The first data access by any desktop AI in a program run opens one blocking
consent dialog. Approving opens access for all desktop AIs for the rest of the
program run; refusing blocks all of them temporarily.
- More than 200 tools expose chronicles, full texts, e-mails,
daily logs, deadlines, master data, whole cases (in parts) and output
formats; documents marked secret are excluded. Changes to master data,
chronicle, calendar, AI memory, full texts and the legal library take effect
only after the user approves them, in an IRONSTICK form,
dialog or the Data Consistency Monitor: master-data and relationship suggestions
land in the monitor; AI-proposed changes to case-bound AI-instruction notes are
presented as a red/green diff and applied only after the user explicitly
accepts; chronicle entries are never written by the AI — it only prefills the
capture form, which the user completes and saves; and an AI-healed document
full text (OCR re-scan) replaces the stored full text only after
the user has seen the complete text in a confirmation dialog and saved it.
Without such a dialog, the AI writes only its own work areas: the encrypted
scratch store, the daily log at the user's request, and its own case
assessment after the deterministic checks. Tags and firm-chat messages need the
user's confirmation in the chat. On a case subscribed from a colleague, the AI
can pin a note that travels through
IRONSTICK SERVER to the case owner and all subscribers
(section 9). Hand-over into the word processor and file delivery are explicit user
steps; navigation only if separately permitted.
- **OCR re-scan healing (local) and the PDF hand-over (user-initiated).**
When a stored full text is truncated or ruined by an old OCR run, the AI may
request a fresh scan: the user picks the pages, the re-scan runs **locally**
(no new external recipient), and both text versions are staged in the local
scratch store; whatever the AI reads from them flows to its provider like any
other tool result. As an absolute last resort — only after such a re-scan and
only within hard bridge-enforced limits —
the AI may ask the user to hand it the **original PDF file**: IRONSTICK
only opens the entry; the user saves the PDF via the paperclip icon
and drags it into the desktop-AI chat. This transfers the document
**as a file** (including stamps, signatures and images) to that AI's
provider — it happens exclusively by the user's own deliberate action, never
automatically, and documents marked secret are excluded from the whole
re-scan channel from the start.
- **Local AI scratch store (data minimisation).** Large tool results can be stored in a local, encrypted scratch file and reworked there **without being transmitted to the AI
provider at all**. Complete case exports for the AI and the one-call
session-start overview are likewise written straight to this local store; only
the portions the AI subsequently reads flow to its provider.
The scratch store lives inside the data
folder and persists across sessions (it is the AI's working context);
files no longer in use are deleted automatically, and no new external recipient is created.
- **The scratch store is encrypted.**
Every file in the scratch store is AES-256 encrypted at rest; files leave the
store only through IRONSTICK's export dialog, never to a download folder.
The database, the document store, the configuration file and the scratch
store are thus all AES-256 encrypted at rest. Temporary working copies of opened
documents (cleared at every start and at shutdown), the short-lived peer-review
buffer, the scan inbox, exports and outgoing files remain in clear text
(section 8).
- Tool results are delivered to the user's desktop-AI client and from there
to **its provider**: Anthropic for Claude (under the user's Claude
subscription and Anthropic's terms), OpenAI for ChatGPT (under the user's
ChatGPT subscription and OpenAI's terms), Alibaba Cloud/Qwen for Qwen
Desktop (under the user's Qwen account and Qwen's terms) or Moonshot AI for Kimi
Desktop (under the user's Kimi account and Moonshot AI's terms). IRONSTICK writes no
audit-log rows for bridge traffic; the top bar shows one pulsing logo per
active AI (Claude mark, ChatGPT mark, Qwen mark, Kimi mark) — simultaneous access is
possible and each is signalled separately. Each bridge program keeps a
**local log** next to itself with time, tool name, success/failure and
answer size, and **never content**. A second, more detailed
**diagnostic log** in the data folder is **off by default**; when
switched on for troubleshooting, it can contain call parameters (which may include names) in clear text and is trimmed automatically. Neither log is transmitted anywhere.
- The shared AI memory stores distilled insights (client- or case-bound) locally. The AI can propose to store, change or delete an
entry; each change takes effect only after the user approves it. Entries linked
to a case are never deleted automatically; entries without a case are deleted
after one year without use. The memory's content is visible to every AI
channel and is transmitted whenever those channels read it.
- Peer review between the desktop AIs runs through IRONSTICK, not directly
between the providers: one AI leaves a work product (e.g. a draft) in a
**local, short-lived exchange buffer** (plain files on the user's machine,
deleted when the task is closed and wiped completely on every app restart) and a
task note; the other AI reads it when the user directs it to. This creates
**no new external recipient** — whatever the reviewing AI reads is transmitted
to **its own** provider exactly as any other tool result above, under that
provider's terms. Providers never exchange data directly; what the reviewing AI
reads goes only to its own provider.

### 4.10 Audio transcription (speaker assignment)

Speech recognition runs entirely locally; the recording itself never leaves the
computer. If the user has specified more than one speaker, the AI then assigns the
recognised lines to the speakers: section by section (medium tier), the recognised text, plus lines already assigned as context and the number of
speakers. The calls run through the router — pseudonymised, subject to consent,
recorded in the audit log. The answer must not alter the text: every line is checked
against the recognised text; if a block fails, the transcript is taken over without
speaker assignment. With only one speaker, no AI call takes place. If free
background providers are switched on (section 1), these calls may run through
them.

## 5. The AI instruction package (export ZIP)

![IRONSTICK Client, user with MD files and ZIP archive, AI web app](https://www.ironstick.legal/img/67_instruction_zip.png)

Building the package is local and involves no AI call. It produces
**clear-text Markdown** (no pseudonymisation) intended to be uploaded by the
user to an external LLM of their choice. Content:

| File | Content |
| --- | --- |
| Cases | all cases of the selected client with parties, status and evidence-ID index (no full texts) |
| Entities | **all persons/entities of the entire practice** (not only this client) including national personal and company ID numbers (e.g. Romanian CNP/CUI), date of birth, full contact data, profession, vehicle — plus the stored analysis blocks (assessment, vulnerabilities, residences, finances, social environment, sources) |
| Institutions, Deadlines | all institutions; active deadlines with case numbers and client name |
| Mail | the 100 most recent e-mails **practice-wide**: sender, recipient, subject, attachment file names, short snippet (no full bodies) |
| Daily log | last 10 days, practice-wide, including the case chronology of that window |
| Per case | one case file per case: the complete chronicle **with document full texts**, client and party identifiers including ID numbers; plus all case background notes uncapped |
| Instruction files | five working-rule documents — no personal data |

**What happens when this ZIP is uploaded to a third-party AI service:**
the user personally transfers, in clear text, substantial parts of the entire
practice — client identities with national ID numbers, third-party profiles
including sensitive assessments, practice-wide correspondence metadata and case
files — to that provider. From that moment the data is processed under that
provider's terms (training use, retention, jurisdiction). IRONSTICK shows a privacy warning before the
export (once per program run) and offers the export also to Google Drive, which
additionally places the files with Google. The user acts as the transmitting
controller under GDPR and must ensure a legal basis (e.g. Art. 6, professional
secrecy rules) before uploading.

Documents marked secret are excluded from deadlines, daily logs and case
files.

## 6. Summary matrix

| Flow | What leaves (form) | Pseudonymised | Own consent dialog | Audit log |
| --- | --- | --- | --- | --- |
| Document capture (4 calls) | OCR and document text (limited), case and name lists | yes | yes — once per program run | yes |
| E-mail triage (manual fetch in the e-mail screen) | sender, subject, short snippet (no case list); only for e-mails not assigned locally | yes | no | yes |
| E-mail triage (background, every five minutes) | same | yes | no | yes |
| Daily briefing (API) | briefing text: deadline counts, court dates with case number, client name, city and court, appointment titles | yes | no (first start of the day, prepared during start-up) | yes |
| Person research, stage 1 | real name (web search) | no — by design | yes — once per session | yes (clear text) |
| Person research, stages 2 and 3 | page texts (limited per source) | yes | same consent | yes |
| Person/institution autofill | typed name (web search) | no — by design | yes — once per session | yes (clear text) |
| Daily-log AI import | pasted raw text | yes | yes | yes |
| Data Consistency Monitor | per new document: its contact regions and names of possible matches (no contact data or ID numbers from master data) | yes (institution names in clear text) | no (throttled background) | yes |
| Word processor: revise / structure | selection with a context excerpt / document lines (limited) | yes | no (button = consent) | no |
| Read-aloud preprocessing | each sentence, clear text | no | no (start = consent) | no |
| Red/Blue Team review (API) | draft text, date and jurisdiction | yes | no (switch = consent) | yes |
| Dialogue summary (API) | older dialogue turns (user questions and final answers only) | yes | no (background pass of the API chat) | yes |
| File→Markdown AI correction | converted text | yes | no | no |
| Redaction AI finder | already-redacted text; answers contain third-party clear names | off — by design | toggle = consent | yes |
| AI chat (external) | prompt, screen context, memory, most recent dialogue with summary, tool results (limited per item) | yes | yes — once per session | yes |
| Desktop-AI bridge (Claude Desktop/Code, ChatGPT Desktop, Qwen Desktop, Kimi Desktop) | raw tool results (full case data) | no — by design | yes — once per program run for all desktop AIs | no (app side) |
| Audio transcription: speaker assignment | recognised text in sections — never the recording | yes | yes | yes |
| Free background providers (Groq, OpenRouter, Mistral) — only when switched on | the same content as the respective background task; never chat, pleading or Red/Blue Team | as the respective task | no (switch per provider = consent) | yes |
| AI instruction ZIP | clear-text practice export (user-initiated upload) | no | warning once per program run | n/a |

## 7. Non-LLM external services (for completeness)

![IRONSTICK Client, internet, location symbol and weather symbol](https://www.ironstick.legal/img/68_external_services.png)

Independent of any LLM, the following features contact external services with
the minimum data needed: number validation (EU VIES, EORI, GLEIF, business
registers — the number being checked), the person search's direct web queries
(the name), IMAP (your mail server), and optional exports to Google Drive (the
exported files). None of these involve the AI provider.

The following services are contacted in the background or with equally
minimal data:

- **OpenStreetMap Nominatim (address geocoding)** — once a day, a
background job works through the persons and companies in the party records
that have an address or a city, and sends street, district, city and country
(never the name) to Nominatim to place them on the map; an address already
placed is sent again only when it changes, and one that was not found is retried
after a waiting period. The place name of the weather lookup is geocoded the
same way. No consent dialog; the coordinates are stored locally.
- **ipwho.is (geo-IP lookup)** — once per program start, as part of the
local security snapshot (section 8), IRONSTICK asks this service for the
machine's **public** IP address and a coarse location (city, country, ISP).
The request carries no payload; the service receives the public IP address from
which the request arrives. The answer is stored exclusively locally and is
transmitted nowhere else.
- **Open-Meteo (weather)** — on request, the AI weather lookup (section
4.8) queries current weather data for a named place; the daily briefing does the
same for the town stored in the configuration when its optional mode with a
weather reference is on. Only the place's coordinates are transmitted — never
case, client or person data.
- **[www.ironstick.legal](https://www.ironstick.legal/index.html) (news file)** — after the access code has been
entered and then every six hours, the application downloads a news file with
notices and warnings from the software vendor's website, plus any image it
references. The request carries no data; like every web server, the vendor's
server sees the IP address and the time of the request.

None of these services involves the AI provider.

## 8. Local storage, integrity and user control

![IRONSTICK Client with TOTP access lock; documents, scratch store and database encrypted with AES-256](https://www.ironstick.legal/img/69_local_encryption.png)

Independent of the AI channels above, the following applies to **all**
data IRONSTICK holds:

- **Exclusively local storage.** The entire data set — database, document
files, media, transcripts, configuration — lives in a single data directory at
a path the user defines. IRONSTICK requires no cloud account,
and no function sends user data to the vendor (the only vendor contact is the
news download, section 7). Where a practice runs
the optional IRONSTICK SERVER (section 9), that server too is the firm's own
device on the firm's own network — nothing about vendor access changes.
- **Database encrypted at rest (AES-256).** The database — the record
of clients, cases, chronicle, e-mails, deadlines, transcripts' text and the
AIs' memory — is stored encrypted with AES-256, every page individually and
with an integrity seal per page. The key is derived by the application at
run time and is never written to disk or configuration, so a database file
copied from the storage device or a backup archive cannot be read without
IRONSTICK. This is a technical measure in the sense of Art. 32 GDPR that
applies without any user configuration and cannot be switched off.
- **Documents encrypted at rest (AES-256).** The document store — the
original PDFs, scans, images, audio and transcript files — is encrypted the
same way, file by file, with an integrity seal that detects alteration.
Reading happens through a single gate that decrypts the requested file
into a temporary working copy and removes those copies when the
application closes or starts. Backups and case packets carry the documents
in their encrypted form.
- **What remains in clear text.** Besides the working copies of the current session and exports: scans waiting in the scan inbox before they are captured;
the short-lived peer-review buffer between desktop AIs (section 4.9, deleted
when the task is closed and wiped at every start); and, only if switched on,
the detailed bridge diagnostic log (section 4.9).
- **Access control at start (TOTP).** Once the licence key is
accepted, the application shows nothing but a code prompt until the user
enters the current six-digit time-based one-time code from an
authenticator app (RFC 6238). The secret behind the codes is never written
to disk; only a
non-secret fingerprint of the confirmed set-up is stored so the QR code is
shown again after a key change. Repeated wrong codes trigger a growing
wait. The desktop-AI bridges receive no case data while the application
is locked. The configuration file (API keys, e-mail credentials,
server access) rests AES-256 encrypted with the same mechanism as the
documents and is decrypted in memory only.
- **User-controlled storage location.** The data path is chosen during
setup and can be changed at any time — including to a removable medium such as
an encrypted USB drive or an encrypted external disk. The application simply
follows the configured location; encrypting the medium (e.g. with the
operating system's disk encryption) is in the user's hands and is
recommended for portable media. Network drives and cloud-synchronised folders
(OneDrive, Google Drive, Dropbox, iCloud, Box, pCloud, MEGA) are refused as data
location. When the storage location is changed in the configuration, IRONSTICK checks the target against these rules; copying is done with checksum comparison, the switch happens only after successful verification, and the old folder is kept under a new name.
- **Device synchronisation without any cloud.** The optional
Windows ↔ Android-tablet synchronisation runs over a direct USB
cable connection — no third-party service is involved, transfers are
checksum-verified, and a mandatory backup is taken before the database is
replaced. Documents marked secret are never transferred to the tablet at
all.
- **Start-up security snapshot (local login log).** At every program
start, IRONSTICK records in the background a snapshot of the machine it is
running on: host name, operating system, RAM/CPU, local IP address, default
browser, running remote-access tools it detects, and a derived security
score — plus, once per start, the public IP address and coarse location
(city, country, ISP) obtained from the geo-IP service ipwho.is (section 7).
The snapshot is stored **exclusively locally** in the database and is transmitted nowhere. Retention is limited
automatically to a maximum of **6 months** and a maximum of **1,000
entries** (daily housekeeping), and the user can inspect the log at any
time in the configurator (card "Login log").
- **Integrity fingerprints on every captured file.** At the moment of
capture, every stored file receives a cryptographic content fingerprint
(SHA-256 for documents and media; a legacy MD5 fingerprint for transcript audio
files), recorded in the database next to the entry. Any later modification of a
stored file is therefore **detectable** by comparing the file against its
recorded fingerprint, and e-discovery exports (EDRM) carry these fingerprints
so that recipients can verify the files independently. This makes any manipulation evident and provable.
- **Backups at the user's discretion.** A built-in backup function writes
a complete copy of the data (full backup or data-only backup) as a single ZIP
archive to any destination the user selects, and restores from such an archive
on demand. How often backups are made, where they are kept, how long they are
retained and whether the archives are additionally encrypted is entirely the
user's decision. Inside every archive, the database, the documents, the
configuration and the AI scratch store are AES-256 encrypted in any case; only
scans still waiting in the scan inbox are included as they are —
IRONSTICK imposes no schedule and transmits backups nowhere.
- **Generated documents belong to the user.** The user can generate
documents at any time — pleadings, letters, dossiers, lists, exports — and
download them: they are written to the local Documents folder or, where the
user explicitly selects Google Drive as a destination, to the user's own Google
Drive account.
Whatever happens to a
generated document afterwards — printing, filing with a court, e-mailing,
uploading to any service, sharing with third parties — is decided by the user alone.

**Allocation of responsibility.** Because IRONSTICK stores
everything locally and transmits data only on the paths documented in sections
1–7 and 9 — including the background functions that run without a dialog of
their own (sections 3 and 7) — the operator of the practice is the sole data
controller: they choose the storage medium and its encryption, manage and
retain backups, decide which AI functions are enabled, give or withhold each
consent, and alone decide what is done with every document, export or backup
the software produces. AI-generated content — drafts, summaries, assessments,
classifications — is always a **proposal** on which the user decides. This document exists so that every
one of those decisions can be made on full knowledge of the actual data
flows.

## 9. IRONSTICK SERVER (optional, on the firm's own network)

![IRONSTICK Server and IRONSTICK Client connected over the LAN; the server has no connection to the internet](https://www.ironstick.legal/img/70_server_lan.png)

Practices with several IRONSTICK workstations can operate the optional
**IRONSTICK SERVER** — a coordination service running on the firm's own
Synology NAS, inside the firm's own local network. For the purposes of this
declaration the decisive facts are:

- **No external party is involved.** The server is the firm's own
device. It accepts connections only from private (local) network addresses
and **never initiates any connection to the internet** — no telemetry, no
update checks, no vendor contact. The software vendor has no access of any
kind.
- **Transport and storage.** Documents travel to the server and rest there
only as AES-256-encrypted files in the same vault format as on the workstation;
the server never sees their plain text. The server runs on the firm's own NAS
inside the firm's network, without internet connections.
- **No AI involvement.** The server carries no AI traffic, stores no AI
key and never communicates with any language-model provider. All AI channels
described in sections 1–6 remain strictly per-workstation; the
pseudonymisation layer, the consent gates and the audit log are unaffected
by the server's presence.
- **What the server stores** — all of it on the NAS, under the firm's
control: (a) **shared case packs** — only cases a user explicitly
released for colleagues; documents marked secret are excluded by design and
never reach the server; (b) the **central legal library** — public
statute texts as Markdown, each with the contributing user's name, source
link and review date; (c) **time-tracking records** — per user, day and
case: case number, client name and minutes, reported by each workstation in
the background; (d) **presence totals** per user and day, derived from
session heartbeats; (e) **case-event metadata** — shares, revocations,
subscription starts/ends and exports, each with date, time and LAN IP
address; (f) a **security log** with administrative and login events,
including LAN IP addresses; (g) **case hand-over parcels** — when a user
transfers case ownership to a colleague, the complete case travels as one
archive through a temporary transfer area on the server; the parcel is deleted from
the server after the recipient has confirmed the import, and documents
marked secret travel only after the owner's explicit additional
confirmation (otherwise the transfer is aborted); (h) the **central
scan intake** (optional Central Scan App): a registry of each user's
**case numbers and display name** (reported by the workstations as
matching material — never case contents) and a per-file **document
spool** holding scanned incoming mail addressed to one user until that
user's workstation has fetched and confirmed it, after which the server
copy is deleted; text recognition and recipient matching happen locally on
the scan workstation, without any AI; (i) **backup event records** —
each workstation reports its data-backup and restore events (kind, file
name, size, time, LAN IP), so the administrator can see per user when the
last backup was taken and remind users who fall behind; (j) the **firm-wide
party directory** — the master records of
**persons and companies, institutions, the relations between them and the
person analyses** kept by the workstations, together with the queue of
change proposals exchanged between users; (k) the **assessments of court decisions** — per workstation and decision: court, panel and composition, the names of the judges and court clerks, decision number, date, court file number, the internal case and client numbers of the assessing workstation, measured text shares and the English analysis of the decision and of both sides’ submissions written by the AI; (l) the **firm chat** — messages between the registered users (deleted after one year) — and notes pinned to a shared case, visible to its owner and all subscribers.
- **The firm-wide party directory.** As soon as a workstation is
connected to the server, its party master data (who is who: names,
identifiers, contact data, relations, person analysis) is **uploaded
automatically** and mirrored to every other workstation of the same firm,
so that each lawyer sees with whom the firm is already dealing and the firm
obtains a firm-wide conflict-of-interest view. Every record has exactly one
owner — the user who brought it in first; all other users hold a read-only
copy and can only send a **change proposal**, which the owner accepts or
rejects in the data monitor. If the owner cannot be reached for 48 hours,
ownership passes to the proposing user. **Case contents are not part of
this directory** — documents, chronology, correspondence and case
assignments remain governed by the explicit share and hand-over actions
described here. The exchange stays inside the firm (one controller), never
leaves the local network and involves **no AI provider**: duplicate
detection and field comparison run as plain program logic on the sending
workstation; the server only stores and forwards. Firms should mention this
internal sharing of party data and person analyses in their record of
processing activities.
- **The assessments of court decisions.** An assessment is stored only on the user’s express command; afterwards, like the party directory, it is **uploaded automatically** and mirrored to every workstation of the firm. Each workstation keeps its own assessment per decision, which only it can change. The analysis describes the content of a decision and of the submissions to it — to that extent it contains **case content** — and names judges and court clerks. No AI provider is involved; the exchange stays within the firm’s local network. Firms should also list this disclosure in their record of processing activities.
- **Deliberate actions only — for case content.** Case content reaches
the server exclusively through an explicit share or hand-over action
confirmed by the user; legal-library texts only after the user accepts the
entry in the data monitor; scanned incoming mail only after the office clerk
has reviewed the document and clicked send to a chosen recipient. The party
directory and the stored assessments of court decisions are the deliberate exceptions: they
synchronise in the background without a per-record confirmation. Time, presence, event and
backup records are operational metadata created by the multi-user operation
itself and are visible to the firm's administrator in the web admin.
- **Controllership.** The firm operating the server remains the sole
data controller. The administrator controls registration approval, can end
subscriptions, delete legal-library entries and uninstall the package —
which leaves the server data directory under the administrator's control on
the NAS.
- **Employee data.** With IRONSTICK SERVER, each
workstation reports its measured working time per day and case (case number,
client name) and every backup and restore event (file, size, time, IP address)
to the server, where the firm's administrator sees it. The firm must inform its
staff accordingly.
- **Client full names in time records.** The time-tracking overview
shows working time per case and per client, which means client names are
stored on the NAS. This stays inside the firm's own infrastructure and its
existing confidentiality regime — but firms should include the NAS in their
technical-and-organisational-measures inventory (access control on the NAS
operating system, volume encryption, backup policy), exactly as they would for
any office file server.

## Annex — Model clause for your client privacy notice (proposal)

The following text is a drafting proposal that a law practice using IRONSTICK
can copy into its own client privacy notice. It is written to be accurate to the
data flows described above. It deliberately avoids citations of statutory articles so it
can be pasted into notices of any structure.

**Before you paste this — checklist for the lawyer:**

1. Have the data-processing agreement of **every** AI provider you have
set up in place, name in the clause only the providers you actually use, and
check the sentence on training and retention against the terms you actually
signed. The clause covers the free background providers (Groq, OpenRouter,
Mistral, section 1) only through the optional sentence in brackets there —
switch them on only if their terms meet your professional-rule requirements,
and delete the sentence if you do not use them.
2. Keep the bracketed sentence about the AI workspace only if you really use
the desktop-AI bridge (Claude Desktop/Code, ChatGPT Desktop, Qwen Desktop
or Kimi Desktop); delete it otherwise.
3. This clause covers the in-app flows described in this document. It does
**not** cover the AI instruction package (section 5) — uploading that export
to any AI service is a separate decision and will normally require the client's
explicit consent.
4. If you have disabled individual AI functions in the configurator, shorten
the list of purposes accordingly — describe exactly what you use, no more.
5. This is a drafting proposal, not legal advice. Have it reviewed against
your bar's professional rules and local data-protection practice before use.

### English (master)

**AI-assisted case processing.** Our practice uses the case-management
software IRONSTICK, which includes AI-assisted functions. For these functions,
selected data from your file may be transmitted to an external AI service
provider (depending on availability: Alibaba Cloud, Moonshot AI, OpenAI or Anthropic, via their APIs), which processes
it on our behalf under a data-processing agreement. We use these functions for:
classifying and registering incoming documents, prioritising incoming e-mail,
checking our records for inconsistencies, drafting and reviewing legal
documents, and legal analysis and research. [Optional — delete if not used: For
tasks that run in the background, such as prioritising incoming e-mail, checking
our records and summaries, the free services of Groq, OpenRouter or Mistral may
additionally be used; we checked their terms before switching them on.]

**Protective measures.** Before anything is transmitted, the software
replaces the direct identifiers stored in our system — names, personal
identification numbers, telephone numbers, e-mail addresses, postal addresses,
case numbers and bank details — with neutral placeholders on our own computer.
The table linking placeholders to real data never leaves our office, and the
provider's answers are translated back locally. Only what the specific function
requires is transmitted, within fixed size limits. A small number of functions
technically require real data — for example public web research on a person's
name, or the read-aloud function; we use these only through a deliberate
individual action. [Optional — delete if not used: For drafting court documents
we additionally use an AI workspace in which case data is processed without
placeholder replacement; this channel is used only under our direct
supervision.] Separately from the AI functions, the software sends postal
addresses stored in our records — without names — to the public map service
OpenStreetMap so that they can be shown on a map.

**The provider.** Under the provider's terms applicable to us, content
transmitted through the API is not used to train AI models
and is stored only for a limited period. Processing may take place outside the
European Economic Area; in that case the transfer is protected by the
safeguards agreed in our data-processing agreement with the provider. What the
provider may do with the data is governed by that agreement — not by the AI's
own discretion.

**Your choice.** Our professional duty of confidentiality remains
unaffected: beyond what is described here, no data leaves our practice. The
processing rests on the mandate you have given us and on our legitimate
interest in handling your case efficiently and accurately. You may object to
the use of the AI-assisted functions at any time; where feasible, we will then
handle your file without them. You also retain all your statutory
data-protection rights, including access, correction, deletion and complaint
to the supervisory authority.

This declaration was derived from the application source code
(pseudonymiser, external router, audit log, and each feature's transmission
path) and reflects the shipped behaviour of version 1.0.7 (as of November 2026). Processing by
the AI provider itself — retention, training use, sub-processors, jurisdiction —
is governed solely by the provider's terms, in their current versions. The
links to each provider's terms and privacy policy are also in the corresponding
section of the AI configuration.

| Provider | Terms and privacy |
| --- | --- |
| Alibaba Cloud (Qwen) | [alibabacloud.com/help/en/legal](https://www.alibabacloud.com/help/en/legal) |
| Moonshot AI (Kimi) | [platform.kimi.ai/docs/agreement](https://platform.kimi.ai/docs/agreement) |
| OpenAI (ChatGPT) | [openai.com/policies](https://openai.com/policies) |
| Anthropic (Claude) | Commercial Terms of Service: [anthropic.com/legal/commercial-terms](https://www.anthropic.com/legal/commercial-terms); Usage Policy: [anthropic.com/legal/aup](https://www.anthropic.com/legal/aup); Privacy Policy: [anthropic.com/legal/privacy](https://www.anthropic.com/legal/privacy) |
| Groq | [groq.com/terms-of-use](https://groq.com/terms-of-use) |
| OpenRouter | [openrouter.ai/terms](https://openrouter.ai/terms) |
| Mistral | [legal.mistral.ai/terms/commercial-terms-of-service](https://legal.mistral.ai/terms/commercial-terms-of-service) |
