For starters, could you briefly explain to us what brighter AI does?

Marian Gläser: brighter AI was founded in 2017, and we are supported by a few investors from the security industry, as well as the Deutsche Bahn.

At brighter AI, we provide image & video anonymization solutions based on state-of-the-art deep learning technology. Our solutions, Precision Blur and Deep Natural Anonymization (DNAT), redact faces and license plates and help companies comply with data protection regulations such as the GDPR. We enable companies in various industries to use publicly-recorded camera data for analytics and AI. With our solution, they can mitigate their liability and the risks of being fined, increase the capacity of their teams, improve their time to market, and push innovation.

We fully enable smart cities, public transportation, autonomous driving, and so on. This is our place in the value chain of data handling. With our solutions we ensure innovation is not stifled and identities are protected.

Why should municipalities or public transport companies care about privacy?

MG: On one hand, there are stringent regulations; in Europe you have the GDPR, in the US you have the CCPA and we can see that more and more countries adopt legislation to protect people’s identities. Regulators understand that with data being collected in public, it becomes critical to maintain the people’s privacy. Companies that violate those regulations can be fined up to 4% of their yearly revenue, and have to make the privacy issue public.

There’s a movement towards the regulators understanding that with data being collected in public, it is critical to maintain the people’s privacy.

- Marian Gläser, Co-Founder/CEO of brighter AI,

On the other hand, of course, there’s the social responsibility. You not only have to make sure that your trains are safe, but also that the privacy of your passengers is secure.

Why do they need anonymisation solutions? What are the most common use cases?

MG: In the case of public transportation, and smart cities in general, you don’t need to have active consent in order to use the data and make heat maps, know where people are going, or shopping, etc. It’s just not feasible to get everyone’s consent. You also can’t do it like the Uber way, where purchasing a ticket means you give the company permission – it wouldn’t fly by the GDPR.

The recital 26 of the GDPR states that anonymous data doesn’t apply to the GDPR anymore. Hence why anonymisation and taking personal identifiable information out of data really change the picture. People are protected and the regulation is also complied with.

We offer a service that anonymises faces and license plates while still making them usable for analytics by public transport companies. What we do is, we replace the face with an artificial identity that still shares the original person’s attributes to keep the data as raw as possible without decreasing the quality : age, gender, ethnicities, emotions, etc. Traditional anonymisation with pixelation is also available for other use cases, but adding artifacts can be cumbersome for the neural network and make the data questionable or unusable.

Adding artifacts [to anonymise data] can be cumbersome for the neural network and make the data questionable or unusable.

- Marian Gläser, Co-Founder/CEO of brighter AI,

Can you give us some examples of what you have done with your existing customers in this industry?

MG: We work with some famous names in the industry, such as the Berlin Public Transport, the Deutsche Bahn… The cases vary depending on the provider. For some, we are anonymizing the data right within the train. That data is then exported to the cloud, where multiple analytics can be run, the cleanliness of the train for example, as well as the numbers of seats occupied.

We also provide bus and tram drivers with study material to learn their daily route and interact with any pedestrians or vehicles that might cross their path. The identification of faces lets them know if a cyclist has seen the upcoming bus, for instance.

Autonomous trains also need this data material to understand the behavior of passengers waiting on the platform when entering the station, and brake when necessary.