Privacy During Field Testing

This post explains how privacy works during field testing. Currently, we are still field testing. We expect to continue field testing through 2022. We will update this section should anything change. To find out how privacy works upon completion of field testing, click here.

During our field testing phase, we are collecting and securely storing more data than we will upon its completion. We use this data to train and optimize our fraud-detection algorithms. Without it, we wouldn’t be able to fairly and inclusively give a share of Worldcoin to as many people as possible on Earth. But we can’t wait to stop collecting it and we want to make it clear that it will never be our business to sell your personal data.

Any individual who consents to our collection and use of their biometric data may revoke their consent at any time and this data will be deleted. We will delete all the biometric data we have collected during field testing once our algorithms are fully-trained (this may be slightly after the conclusion of field testing).

Data We Are Collecting

We collect the following data through the Orb after the user gives us their consent:

1. Images of users’ body, face, and eyes, including users’ irises (visible, near infrared and far infrared spectrum)
2. Three-dimensional mapping of users’ body and face

Why We Are Collecting The Above

We process the above data exclusively for the following purposes:

1. Calculating IrisHashes
2. Comparing a prospective user’s IrisHash with other IrisHashes
3. Optimizing how the IrisHash is calculated
4. Training and developing neural networks to be capable of recognizing human irises
5. Removing bias from neural networks to make sure our systems work equally well with every user
6. Training and developing a fraud-detection system that is capable of detecting whether a user is a real human being or not
7. Training and developing models to create artificial iris images to use as additional training data for our algorithms
8. Training and evaluating personnel working on implementing and optimizing the above systems

More On Why

In order to fairly and inclusively give a share of Worldcoin to everyone on earth, the system for claiming a free share has to work flawlessly, treat everyone equally and, importantly, cannot be susceptible to fraud. After all, we don’t want anyone to be able to claim their share more than once. But this means that our Orbs and the algorithms that run on them have to be nearly perfect; in order to achieve near perfection, we have to train and optimize our algorithms. That requires a lot of data. The more data we collect now, the more easily and accurately our algorithms will be able to differentiate between people and the sooner we can stop collecting biometric data altogether.

More specifically, the algorithms running on the Orb have to be able to:

1. Distinguish between minuscule differences in human irises

To do this, we use machine learning. This significantly improves algorithm performance but also requires a lot of pictures of human irises.

1. Detect a user’s eyes in relation to the rest of their face and, once the eyes are located, identify the irises

This allows the user to claim their free share.

1. Detect whether a user is a real human being or not

For example, we capture a short video of each user to confirm that they are indeed alive and human (rather than a display, a bot or an animal). This increases the security of our system.

1. Run without bias

We use the data we collect to identify and address any potential biases that currently exist in the system. For example, we routinely make sure that our algorithms are performing equally well among different races, ages, genders and ethnicities.

We can’t wait to finish field testing and stop collecting biometric data. We really can’t. But now you understand why we have to do it right now: it’s the only way that the Worldcoin project will work.