Introducing Velum: Privacy-first AI masking
Every time you paste sensitive information into ChatGPT, Claude, or any other LLM, you are making a trade. You get the answer you need, but you give away data that was never meant to leave your organization. Names, addresses, financial records, proprietary notes. All of it becomes part of someone else's training pipeline.
We built Velum because we believe that trade is unnecessary.
The problem is architectural
Most privacy solutions in the AI space are afterthoughts. They bolt encryption or access controls onto systems that were never designed to keep data local. The result is a compromise: either you sacrifice functionality for security, or you accept that your data will be processed on someone else's servers.
Velum takes a different approach. Instead of trying to secure the cloud, we eliminate the need for it entirely. Velum runs entirely on your device. Your data never leaves your machine. Not for processing, not for storage, not for analytics.
How it works
The concept is simple, but the execution is nuanced. When you input text into Velum, it automatically identifies personally identifiable information (PII): names, emails, phone numbers, identification numbers, financial data. It replaces each instance with a reversible token.
The masked text is what gets sent to the LLM. The model processes it normally, because the structure and context of the text remain intact. When the response comes back, Velum restores the original values, giving you a complete, accurate answer without ever exposing sensitive data.
The goal is not to make the cloud safer. The goal is to make the cloud irrelevant for sensitive data.
Why local matters
There is a fundamental difference between "we encrypt your data" and "we never see your data." Encryption protects data in transit and at rest, but at some point, someone holds the keys. Local processing removes that liability entirely.
For organizations in regulated industries like healthcare, finance, and legal, this is not a preference. It is a requirement. HIPAA, GDPR, and industry-specific regulations all impose strict limits on where personal data can be processed. Velum makes compliance straightforward by design.
The broader vision
Velum is the first product in the Adimen Suite, but it reflects a principle that runs through everything we build: privacy should be the default, not an add-on. We are not interested in collecting data about how you use our tools. We are not building profiles or training models on your inputs. We are building infrastructure that treats your data as yours.
The future of AI is not about better cloud services. It is about bringing intelligence to where the data already lives. Velum is our step in that direction.
See how Velum masks your data before it reaches any AI, and explore the use cases for where it fits your work.