When you use PQAI for running prior-art searches, you can rest assured that PQAI provides you complete privacy.

Unlike most search engines, which track everything you do, PQAI never tracks or saves your search data. We believe that it is very much needed for a platform like PQAI, which is used by many inventors to validate the novelty of their ideas.

When you enter a search query on PQAI, it goes to our server in the cloud on a secure, encrypted link. The server finds the results matching your query from its database, and sends them off back your way. After this, no traces of your query are left on the server.

(This policy of never storing user search queries is also mentioned on PQAI’s search page – see the link at the bottom of the page.)

Please note even though we don’t track user data, we do store few anonymous traffic statistics such as number of requests. This helps us scale our servers appropriately to handle the traffic, deter abuse, and understand how people find value on our platform.

How do we train our AI?

Another question is whether our AI learns from user behavior? The answer is – no. The fact that we don’t track or save search data makes it impossible for us to train our AI on it.

But that leads to another question: how do we train it then? The answer is: patent office examination data.

We download the examination data that is routinely published by the USPTO on their website, then we process it to create training datasets for our AI. Many contributors from the open source community have helped us in this process.

How do we maintain transparency?

Being part of an open-source effort aimed at accelerating the innovation in patent-mining space, we are committed to also provide the output of these efforts to the community. In fact, we have recently published one of our training dataset on Huggingface. You can find it here.

And it’s not just the dataset, but our AI model is also publicly available. Same goes for the source code of the search engine, which you can find on Github.

Net-net, we maintain 100% user privacy and the open-source nature of our project ensures full transparency.