# ShinrAI > ShinrAI is Innovius’s semantic encryption and PII protection layer for AI applications. It replaces supported sensitive details with realistic alternatives that keep text useful. Your application controls the mapping and can restore originals locally. Use the managed API hosted on STACKIT in Germany or deploy on your own infrastructure, including signed offline activation. ShinrAI serves companies automating customer work and teams using AI with internal files, email and CRM context. For teams moving from Azure text PII, it combines a documented API migration path with reversible replacement, German hosting, local deployment options and transparent record pricing. Keep useful context flowing to AI while controlling which personal details leave your application. Apply protection when indexing files, retrieving context and submitting new prompts. ShinrAI 1.5 is the model of the managed API. Compatibility: a request that names shinrai-v1.4 is answered by ShinrAI 1.5. Pin shinrai-v1.5 in your API requests to stay on this release. Use the latest alias to follow new releases automatically. The managed API offers the current release. A new release is announced in this changelog with its measured changes. Displayed quantities use dots for thousands and commas for decimal fractions; API JSON retains its standard numeric syntax. API usage counts each document separately in 1.000-code-point blocks: Standard ×1, Batch ×0,5, Real-Time ×1,6. Monthly allowances reset on the billing date; purchased packs do not expire. Onboarding is optional, charged once, and can be removed before checkout. Dashboard provides keys, usage and Stripe billing access. GDPR and EU AI Act responsibilities still require application and organisational controls; pseudonymisation is not automatic anonymisation or certification. Secure AI Suite has a published integration guide. Onyx and Langdock integrations are coming soon. Microsoft 365 Copilot integration requires a supported custom-agent or extensibility path; there is no tenant-wide base-URL setting for transparent interception. The developer guides identify supported API contracts and integration steps. Synchronous API content and replacement mappings are processed in memory, without a request-content archive or model training. We keep content only while we process it: for a request, or for the life of a job or session that the customer starts. A request that names a server-held session adds its new replacements to that session. We have no interest in customer data and do not use it for model training or any purpose of our own. A job source is deleted when the run ends. An upload is deleted when the last job that reads it ends; with keep=true it stays 24 hours, extended by every job that reads it. Job results are deleted after at most 24 hours. Uploads, job sources and job results of the native API v2 are encrypted with a separate key per customer; job requests, other jobs and sessions use authenticated encryption bound to the job or session and its owner. Server-held API sessions end after one hour by default and at most 24 hours after creation, or 7 days when support enables extended sessions for the account. Saved templates and dictionaries are the exception to request-, job- and session-bound retention: they remain encrypted until the customer deletes them and are included in operational backups; backups contain no request content, job data or sessions. Local operational records support accounts, key verification, usage and billing reconciliation, without content. Stripe and RevenueCat receive purchase and consumption information, not routine inference text. [Data handling and retention](https://getshinrai.com/docs/data-handling) lists what is kept, for how long, where and who can read it; the privacy notice and data processing agreement explain retention, providers and responsibilities. ## Product and buying guide - [Compare ShinrAI](https://getshinrai.com/public-docs/compare.md): Source-linked comparison with Azure PII, Google, AWS, Presidio, Limina and Tonic; feature availability and measured quality are separate. - [Model languages and categories](https://getshinrai.com/public-docs/coverage.md): ShinrAI 1.5 supports 15 model locales, including Hebrew beta, with 21 categories and 29 subtype classes; API and website language support are distinct. - [Model changelog](https://getshinrai.com/public-docs/changelog.md): 1.5 is the model of the managed API since 2026-10-01; compatibility: a request that names 1.4 is answered by 1.5. API 2.0.0 is stable since 2026-09-29. - [Versioned public evidence](https://getshinrai.com/public-docs/evidence.v1.json): Reviewed model inventory, release status, benchmark identifiers, methodology and source URLs. - [How ShinrAI works](https://getshinrai.com/public-docs/overview.md): Use cases, deployment choices, differentiators and integration boundaries. - [Plans, packs and optional services](https://getshinrai.com/public-docs/pricing.md): Current EUR prices, allowances and one-time onboarding. - [What is a record?](https://getshinrai.com/public-docs/records.md): Per-document rounding, processing weights and a company-chatbot example. - [Interactive estimator and checkout](https://getshinrai.com/#estimate): Adjust workload assumptions, compare plans and review your selection. - [Microsoft Presidio alternative](https://getshinrai.com/alternatives/microsoft-presidio): Where Presidio needs assembly, what ShinrAI adds, migration steps and FAQ. - [Azure AI Language PII alternative](https://getshinrai.com/alternatives/azure-ai-language-pii): Compatible text endpoint at https://azure.api.getshinrai.com, hosting in Germany, restoration mapping, migration steps and FAQ. - [Anonymisierungssoftware für KI-Workflows](https://getshinrai.com/de/anonymisierungssoftware) (German): Pseudonymisation under GDPR Art. 4 No. 5, masking versus replacement, sector use, prices. - [Dokumente anonymisieren mit KI](https://getshinrai.com/de/dokumente-anonymisieren-ki) (German): Document jobs for PDF and DOCX with the redacted PDF back, Nextcloud and n8n integrations, data handling. ## Developer documentation - [API quickstart](https://getshinrai.com/public-docs/quickstart.md): Authentication, a runnable synthetic example and integration links. - [PII API v2](https://getshinrai.com/public-docs/pii-api-v2.md): One stable contract (version 2.0.0) for the hosted API and in-cluster installations; the offline edition serves the native API v1 until its image includes v2. Changes within 2.x are additive; a breaking change gets a new major version and path prefix, announced 12 months ahead. Starts with "Features at a glance": every feature with one line and a minimal curl request. Detect and protect text, tables, JSON, transcripts, pages, images (OCR with pixel boxes, the filled image) and audio (time intervals, the bleeped WAV, at least 10 records per started minute; words the speech recognition mishears stay audible); presets and per-type actions (surrogate, label, mask, partial, generalize, replace, remove, keep); annotations and linkage risk; restore, restore tables, replacements and sessions (24 hours from creation, up to 7 days with the account setting); jobs for text batches, PDF/DOCX documents (the redacted PDF and the protected text) and recordings up to 60 minutes (uploads up to 50 MB, deleted when their last job ends unless kept); limits per tier; errors; capabilities, types and usage. - [API documentation](https://getshinrai.com/docs): Current contracts, limits, Azure migration and offline activation. - Compatibility APIs (for migration; the vendor contract limits types, confidence, media, sessions and jobs, so use the native PII API v2 for full quality): [Azure AI Language PII](https://getshinrai.com/docs/providers/azure) at https://azure.api.getshinrai.com, [Google Cloud DLP](https://getshinrai.com/docs/providers/google) at https://google.api.getshinrai.com, [AWS Comprehend](https://getshinrai.com/docs/providers/aws) at https://aws.api.getshinrai.com. Only the endpoint changes for an existing client; each host serves its own /openapi.json. The same adapters run under /v1/azure, /v1/google and /v1/aws on https://api.getshinrai.com. - [OpenAPI schema](https://getshinrai.com/openapi.json): Machine-readable managed API contract (native PII API v2; vendor adapters under /v1/). - [API explorer](https://getshinrai.com/docs/api#/Native%20PII%20API%20v2): Swagger UI for every contract, opened at the native PII API v2 (recommended), plus the Azure, AWS and Google compatibility APIs under /v1/. - [Native PII API v2 guide](https://getshinrai.com/docs/guides/v2): Choose an input kind, protection action, tier, session or job; curl examples. - [PII API v2: features at a glance](https://getshinrai.com/docs/guides/v2#features): Every v2 feature with a one-line explanation and a minimal request, with the Azure, AWS and Google equivalents where they exist. - [PII API v2 OpenAPI 3.1](https://api.getshinrai.com/v2/openapi.json): The v2 contract with a feature map, a description and examples for every operation, audio included. - [Secure AI Suite integration](https://getshinrai.com/integrations/secure-ai-suite): Available application integration guide. - [Complete public agent summary](https://getshinrai.com/llms-full.txt): These public guides combined in one file. ## Optional - [Partners and support](https://getshinrai.com/#partners): Certified implementation partners and technology enablers; contact info@innovius.ai. - [Data handling and retention](https://getshinrai.com/docs/data-handling): What the service keeps, for how long, where, who can read it and what the customer controls; no model training on customer data. - [Privacy](https://getshinrai.com/legal/datenschutz): Data handling and essential preferences. - [Terms](https://getshinrai.com/legal/agb): Commercial service terms. - [Data processing agreement](https://getshinrai.com/legal/avv): Article 28 GDPR, security measures, subprocessors, return and deletion. - [Imprint](https://getshinrai.com/legal/impressum): Company identity, contact, registration and VAT information. - [ShinrAI 1.3 open weights](https://huggingface.co/innovius/shinrai-pii-m-v1.3): The previous model generation on Hugging Face under the Innovius Open License v1.0 ([licence summary](https://innovius.ai/shinrai/open-license.html)). Free for research, evaluation, teaching and personal use, for public bodies and education, and for companies under USD 10M revenue with "Built with ShinrAI" attribution; not for training competing PII models; Apache 2.0 from 1 September 2028. Free Hugging Face account, automatic approval. The managed API serves ShinrAI 1.5.