BLUEGUARD
BlueGuard Moderator ensures zero data leaks to your
LLMs |ChatGPT | APIs | 3rd Party Systems
BlueGuard is an advanced privacy solution designed for businesses seeking to harness the power of AI, LLMs, ChatGPTs, Privately-hosted models without compromising privacy. BlueGuard enables your existing AI enablements & solutions to work safely, securely while upholding the utmost standards of data security, privacy & compliance (GDPR, DPDPA and more). The platform accurately detects personal data of your customers, redacts them, replaces them with real fake, synthetic data thats contextually aligned with your business.
We are committed to building & improving our AI models to be accurate & better than the industries’ open source or leading cloud providers.
A Security platform for every vertical,
every business
- Your customers account & transaction data whether in structured or unstructured, on files or in-transit needs to be protected
- HIPAA or DISHA are compliances that needs protection of your patient’s medical data
- Your employee, policies information cannot leak into LLMs or 3rd party systems
- Blueguard acts as a shield protecting your platform
- Secure your proprietary information, intellectual property, glossary of terms by providing your custom entities. These are critical to your brand value & any compromise leads to equity damage.
- Finance
Mr PREFIX_1 John PERSON_1 initiated a wire transfer through GooglePay on 23-08-2024 DATE_1. It was sent to a wrong account and the no is 584576581243 ACCOUNTNUMBER_1 Health
- Health
Mr PREFIX_1 Rahul FIRSTNAME_1 had complained of severe heart pain on 02-02-2024 DATE_1 at the AIMS hospital, Delhi STATE_1
- HR
Mrs PREFIX_1 Sharada FIRSTNAME_1 has applied for a loan extension worth $ CURRENCYSYMBOL_1 10,000. Her age is 50 AGE_1 and the risks are high considering her current liabilities
- Insurance
Overall market investment is about$ CURRENCYSYMBOL 500,000 for projectMilShakti CONFIDENTIAL_1 this year will have impacts on team nebula CONFIDENTIAL_2 . Its a value differentiator for Acme ORG_1
- Proprietary & Confidential Info - for all Enterprises
from blueguard_ client import client
from blueguard types import mode
bgx client = client (url= "http://localhost:8080", API_KEY= "<your api key>")
text_ request = "Hi John, how are you ?"
replaced = bgx_client. process._text (text_request, mode. REPLACE)
# now pass the replaced fake data to 11m
from langchain_ openai import OpenAI
11m = OpenAI (temperature=0.2)
response = 11m (replaced. processed_text) # Hi Alex, how are you?
11m = print (response. processed_text) # Hi Alex! I'm fine. Are you well ?
* with mode MASK - Hi PERSON_1, how are you?
# 11m response - Hi PERSON_1, I'm fine. Are you well ?
Customizable Platform Kit - Plug
What you need
- Self-Host or Cloud Saas - Host the platform in your infra as a single container or use our Cloud SaaS API endpoint. We don’t see your data.
- Client-SDK: A simple client-sdk to integrate in your workflows
- API Gateway: A wrapped interface providing a gateway API to public LLMs
- Chat-Interface: A customisable, unified, scalable chat interface that works with your proprietary models or public LLMs
- Admin Portal: Single view for monitoring, alerts, settings and much more.
How We Do It
We “Built For India” & eventually “Build for the World From India”
Our core privacy technology is divided into 2 key parts.
1. A state-of-the-art efficient Machine Learning Model to detect privacy
2. Curated High Quality Dataset with contextual understanding
- We identified 100+ contexts that are diverse & complex, various interaction styles with sentence coherence
- We have a variant of 15+ entities with abstracted ontology of labels including Aadhar, PAN
- We have another variant of a granular 50+ entities including PII, Card Info (PCI)
- De-identify sensitive data, Block your prompts on sensitive data leak, Replace with synthetic data which is contextually aligning with your business or Mask them
- Gain insights into prompt history of your users.
- Hindi, Regional Languages
- Multilingual global languages - English, Spanish, German, French, Italian
Dataset
Our approach is robust, accurate identification of PII, PCI labels in varying diverse contexts.
Machine Learning Model
Our roadmap is to build models & datasets for
We are committed to Open Source and watch this space as we will release the largest Privacy Dataset for India.
We are open to Sponsorship for larger benefits of the ecosystem.
Why Blueguard ?
- We strive to be the leader in detecting personal & sensitive entities accurately by building contextually aware ML models
- We are committed to prevent personal, confidential, sensitive data leaks form your org, so that you can focus on your core critical business use cases & products.
- We continuously evolve & update our models to incorporate the ever expanding growing entities, linguistic context, regional sensitivity
