Our approach to AI

StartMail was built to protect your privacy. Where AI offers new opportunities to make privacy products better, safer and more user-friendly, it is our responsibility to evaluate if and how we can safely implement artificial intelligence in StartMail’s day-to-day operation.

In summary: We do not give AI access to your email communication or private data.

We see possible considerable societal downsides associated with long-term and widespread use of AI. However, we also recognize that over the past year, the technology has developed at an incredible pace, and we would like to present to you our view of whether and how AI can be used within StartMail. We see opportunities to use the potential of AI to build better and safer products for our users. How this can be done responsibly requires an in-depth assessment of the context. Our privacy principles will always guide our decisions on the use of AI.

For a smaller technology company like StartMail, AI can expand what our team is able to achieve. It can help us to improve our users’ experience and develop new privacy tools. Narrowing the feature gap with Big Tech ecosystems is a must to have a real impact on privacy. Working on how to adopt AI thoughtfully to achieve this is part of delivering on our mission.

AI is a fast-developing story, but this is where we currently stand: We are starting to use and test AI in parts of our development work. Our developers, who built StartMail from the ground up and understand its architecture, lead this process. Their human expertise helps us to decide where AI can be useful and how to evaluate its output. They remain responsible for the code that’s ready for release. Using AI in parts of our product development will not give it access to your inbox.

“Move fast and break things” is not our style. Faster development must always meet high security standards. AI-assisted work requires testing and security checks appropriate to each change. More sensitive components must undergo even stricter scrutiny. Our team’s ability to question any approach, identify risks and exercise professional human judgment remains key. It’s an essential element of how we want to adopt AI responsibly.  

AI is handing bad actors more powerful tools than ever before. It would be foolish not to use AI ourselves to test, evaluate and harden our own systems. Where and how AI is allowed to operate matters here. Locally operated models can help examine sensitive code on hardware that StartMail controls without sending it to an external provider. It can help us to scan code, identify vulnerabilities and strengthen our platform.

We are proud to offer 100% human support to our customers. We run no AI chatbots, but we do allow AI to help our support staff work more efficiently, for example by helping update our support articles.

In all this, our commitment to our customers stays clear:

Your emails, attachments, contacts, account information and other private user data will remain 100% off-limits to AI.

Privacy. It's not just our policy - it's our mission.

This statement was written entirely by our team, without AI. Checked with Pangram: 100% human-written.