Our Approach to Data Privacy Ethics in AI Development

All posts

A few years ago, we contributed our knowledge and experience to a project concerning the ethical use of AI by the World Economic Forum (WEF). We shared the ways in which CUJO AI governs its algorithms and set out some suggestions on how AI development should treat private data in an ethical way. This is a short overview that presents our approach to data privacy in AI development through three key aspects.

Learn more about how CUJO AI contributes to the World Economic Forum as its member.

Define Your Goals

The first step we advise AI developers to take is to clearly define what they are trying to build. When you know your goal, it is easier to govern development processes in a way that protects data privacy. Otherwise, you might face a situation where people in your company use private data freely for dubious or undefined aims. There should be no unnecessary use of data in any organization. Governing the development of algorithms is a structured and clearly defined process, where responsibilities, aims, and access can and should be clearly defined.

Governing the Development of Algorithms in Full Compliance

CUJO AI is fully compliant with the strictest data privacy measures, such as ISO-27001, SOC 2 Type 2, GDPR, and CCPA. Do not take shortcuts where private data management is involved, even if it means that you need to put in more engineering into the way data is gathered, stored, and accessed. Anonymization, pseudonymization, as well as other data privacy protection techniques are well known and fully implementable at any scale, as evidenced by our own management of data about 1+ billion devices around the world. Again, this governance of data is made easier when you start with precise goals in mind, as they pave the way for clearer process and team management when it comes to data access.

Process and Governance for Teams

Lastly, we consider machine learning algorithms themselves a key factor in the ethical use of data. What this implies is that AI developers and data scientists should demonstrate and thoroughly test their approaches before using or training models in production. This is a key step many AI developers forget: testing one’s initial results, demonstrating reliability is crucial for machine learning algorithms that will impact real people and might process their data. Ensure that your teams and process governance adhere to a high ethical standard by making sure your approach is actually achieving the goal you’d set out at the beginning, and doing so in a fully compliant, reliable way.

To learn more about the insights we share with the WEF, visit the page dedicated to our contributions to various initiatives at the World Economic Forum in cybersecurity, AI governance, and data privacy.

Want to join us? See career opportunities at CUJO AI.

Other posts by Santeri Kangas

Artificial Intelligence Cybersecurity ISP Security