inlaut for Legal & Healthcare: Local Dictation with True Privilege

Attorneys, notary offices, tax advisors, and physicians share the same dilemma: dictation saves hours every day, but cloud-based speech recognition compromises attorney-client privilege, medical confidentiality, and GDPR compliance. inlaut provides the answer: 100% on-device speech-to-text with zero cloud dependencies, no accounts, and the long-term sustainability of open source.

Note: This page describes the technical architecture of inlaut. It does not constitute legal counsel. Compliance evaluations rest with your organization’s legal or data protection officers.

AI-generated

The Dilemma of Cloud Dictation in Confidential Fields

When dictating client counsel, court briefs, or sensitive medical records (GDPR Art. 9 / HIPAA), transmitting unencrypted voice audio to remote cloud servers carries severe compliance risks:

• Legal Privilege & Professional Secrecy: Disclosing confidential client matters or health records to third-party cloud infrastructure creates substantial regulatory and liability vulnerabilities.
• Special Category Data (GDPR Art. 9): Medical histories, diagnoses, and litigation files demand strict protections that generic AI SaaS providers rarely fulfill.
• Permanent Audio Retention: Many consumer dictation apps default to saving raw audio files indefinitely on remote servers or unencrypted local directories.

How inlaut Protects Confidentiality by Design

inlaut was engineered from day one to guarantee that confidential voice audio never leaves your local workstation:

• 100% On-Device Processing: The neural speech model (Parakeet Ultra) executes entirely on local hardware (utilizing Apple silicon’s Neural Engine on Mac).
• Zero Cloud Intermediaries: inlaut runs no transcription servers. Speech recognition functions fully without an active internet connection—even inside strictly air-gapped practice networks.
• No Audio History or Retention: The microphone is only energized while holding your designated push-to-talk key. Audio samples are held strictly in transient RAM during recognition and discarded immediately afterwards. No audio logs or recording archives are kept on disk.
• No DPA Required: Because no voice or transcript data is transmitted externally, no Data Processing Agreement under GDPR Art. 28 is required for dictation.

Future-Proof Independence Through Open Source

For legal practices and clinics, software vendor lock-in represents a critical operational risk. Many firms have experienced legacy vendors acquiring established tools, inflating license prices, or forcing unwanted cloud migrations.

• Zero Vendor Lock-in: inlaut is licensed under the GNU General Public License (GPL-3.0). The codebase belongs to no single corporate entity and remains permanently open.
• Complete Auditability: IT security teams, data protection officers, and external auditors can inspect every line of code on GitHub. There are no proprietary black boxes, hidden telemetry beacons, or unauthorized model training on your text.
• Perpetual Operational Continuity: Even if upstream development were ever discontinued, your internal IT or third-party service provider can fork, adapt, and run inlaut indefinitely.
• Free Instead of Four-Figure Licenses: While traditional professional dictation packages (such as Dragon Legal or Dragon Medical) cost upwards of $800 to $2,500+ per seat, inlaut is 100% free forever—even for commercial firm and hospital deployments.

Seamless Practice Workflow Integration

inlaut adapts directly to your established desktop environment without disrupting existing tooling:

• Types Directly at the Cursor: Whether drafting briefs in Word, composing emails in Outlook, entering case notes in practice management suites, or filing EHR entries: inlaut pastes recognized text directly into whichever application has focus.
• Custom Replacements for Terminology: Define custom substitutions for statutory abbreviations (e.g. GDPR, HIPAA, USC, CFR), Latin phrases, or complex medical diagnoses.
• Voice-Triggered Snippets: Insert multi-line disclaimers, standardized clauses, or firm signatures with a single spoken trigger phrase. Custom dictionaries and snippets can be exported and distributed across your team.

Platforms: macOS and Windows in Development

inlaut currently supports any Mac with Apple silicon running macOS 26 or later. Recognizing that a vast portion of medical and legal workstations worldwide run on Windows, a native Windows version is in active development, maintaining the exact same uncompromising zero-cloud privacy architecture and open-source foundation.

FAQ for Practice & Clinical IT Teams

Does inlaut initiate outbound network connections?

Only three strictly scoped requests: the one-time speech model download from Hugging Face during setup, the daily update feed check from inlaut.de/updates/appcast.xml (which can be disabled entirely in app settings), and update package downloads from GitHub. Spoken audio and text are never part of any network payload.

Can inlaut be deployed on isolated or air-gapped networks?

Yes. Once the initial speech model file is installed, inlaut operates completely offline without needing any internet connection.

How does the clipboard handle privileged text?

inlaut inserts text via the clipboard at the current cursor position. The clipboard payload is marked with transient flags so privacy-conscious clipboard managers do not persist it, and inlaut immediately restores your previous clipboard contents.

Are enterprise support contracts or SLAs available?

inlaut is an independent open-source project and does not offer paid SLAs. Detailed specifications for security and network architecture are documented openly on our inlaut at work page.

Inquiries from practice IT or compliance officers? Reach out to hallo@inlaut.de.