English | Nederlands

AI Ready Guard 1.0

AI Ready Guard is a local-first Windows privacy assistant for technical clipboard text. It helps DBAs and other technical professionals detect and consistently anonymize potentially sensitive text before sharing it with an AI tool.

The user always makes the final decision. AI Ready Guard never automatically blocks, pastes, uploads, or replaces content.

AI Ready Guard uses best-effort detection. It is not enterprise DLP, endpoint security, monitoring, or an automatic security guarantee. Always review the anonymized result before sharing it.

Copy → Review → Anonymize → Share

Supported in Version 1.0

The supported production workflow is clipboard text:

Clipboard text
      ↓
Local deterministic analysis
      ↓
Detection and explainable risk assessment
      ↓
Consistent local anonymization
      ↓
User review and explicit copy

• Safe, completely analyzed clipboard text remains silent.
• Potential risk opens a small modeless RiskCard without taking focus.
• The anonymized preview appears first; showing the original is explicit.
• At most three plain-language reasons explain why the card appeared.
• One click copies the complete anonymized text.
• DEGRADED and FAILED analyses are never presented as safe.
• Test, Internal, and Production profiles support different privacy boundaries.
• English and Dutch can be switched immediately from the tray menu.

Screenshot analysis — Experimental Preview

Version 1.0 retains the local screenshot extraction architecture as an Experimental Preview. It is not part of the supported protection promise.

• Recognition can omit or alter identifiers, punctuation, operators, paths, addresses, or complete lines.
• Every recognized result remains incomplete and requires comparison with the source screenshot.
• Only recognized text can be reviewed and anonymized.
• The source image is not masked, protected, saved, or changed.
• Do not treat anonymized recognized text as proof that the screenshot is safe to share.

Screenshot OCR and masking research is frozen for Version 1.0. Further work belongs to the separately validated Version 2.x roadmap.

Detection and context preservation

AI Ready Guard detects established sensitive categories such as secrets, passwords, API keys, tokens, labeled personal data, usernames, network addresses, internal domains, connection components, database objects, and file paths. Context-bound rules cover common Oracle, PostgreSQL, SQL Server, Linux, SSH, Git, Docker, and UNC forms.

Anonymization replaces detected components rather than complete lines. SQL structure, delimiters, connection-string keys, plan operators, ports, file extensions, and other diagnostic context remain visible where policy permits. Equal values receive the same replacement token within one analysis.

The preview is limited to 20,000 characters or 200 lines for responsiveness. The copy action still uses the complete anonymized text.

Install

Download AIReadyGuard-1.0.0-Setup.exe from the v1.0.0 release and run it. The per-user setup installs in %LOCALAPPDATA%\AIReadyGuard, creates desktop and Start menu shortcuts, and requires neither administrator rights nor Python.

The setup and executable are not digitally signed. Windows SmartScreen may therefore show a warning. Verify the published SHA-256 checksums before running the package.

See the complete Installation Guide. A portable ZIP and Python wheel are also provided.

Daily use

1. Start AI Ready Guard from the desktop or Start menu.
2. Select Test, Internal, or Production from the tray menu.
3. Copy the technical text you intend to share.
4. If a RiskCard appears, review the anonymized preview and reasons.
5. Choose Keep original or Copy anonymized version.
6. Review once more in the destination before sending.

See the User Guide for profiles, status meanings, examples, and privacy boundaries.

Privacy and offline behavior

AI Ready Guard itself does not transmit or persist clipboard text or images. Current content is processed temporarily in process memory. Runtime source has no network client, telemetry, content log, account, cloud model, or update service.

This does not control Windows Clipboard History, clipboard sync, pagefiles, crash dumps, security tools, other processes, or the destination application. Python and Qt also cannot guarantee immediate physical-memory erasure.

See PRIVACY.md for the exact claim and boundaries.

License

AI Ready Guard project code and documentation are available under the MIT License. The separately packaged PySide6, Shiboken6, and Qt libraries remain under LGPLv3; see the Qt LGPL notice and complete texts in LICENSES.

Development and verification

AI Ready Guard requires Python 3.12 x64 on Windows. The deterministic clipboard engine uses PySide6 and regex; the Experimental Preview also uses pinned local Windows Runtime OCR projections. Runtime performs no dependency download.

py -3.12 -m venv .venv
.venv\Scripts\python.exe -m pip install .
$env:PYTHONPATH = "src"
$env:QT_QPA_PLATFORM = "offscreen"
.venv\Scripts\python.exe -m unittest discover -s tests -v
.venv\Scripts\python.exe tests\ui_smoke.py

See the Developer Guide, Architecture, Version 1.0 Release Notes, the Version 1.0 Quality Report, the Packaging Verification, and the Final Checklist.

Deliberately outside Version 1.0

No dependable screenshot protection, image masking, finding-to-pixel mapping, local LLM, cloud model, OCR engine voting, PDF processing, browser extension, keyboard hook, active-application detection, account, enterprise policy, dashboard, telemetry, or automatic update mechanism is included.

The Golden Screenshot Corpus, renderer, OCR reports, capture study, and readiness review remain in the repository as engineering evidence. They do not expand the Version 1.0 support promise.