Port87 does not use your email to train AI. Port87 also does not use AI to read your email. We do not use AI in the development of Port87. Port87’s code is written by real people, and much of it is open source.
We do not use AI for a number of reasons:
- Training AI on private email content can leak private information. It is possible to craft a prompt to retrieve verbatim copies of real emails used to train them.
- Using AI to read email can introduce security vulnerabilities. Agentic AI can be targeted using a technique called “prompt injection” to divulge private information to malicious third parties.
- AI is prone to hallucinations and inaccurate results.
- AI generated code is unreliable, unmaintainable, and prone to security vulnerabilities.
- AI generated code may infringe on the copyrights of the developers whose code was used to train the AI. In fact, testing has shown that AI generated code has already infringed on our copyrights.
- AI does not provide enough value to justify its astronomical cost, both monetary and environmental. In fact, we struggle to see any value AI provides, considering the real threat of security vulnerabilities and inaccuracies.
We value your privacy, and we do not and will not mess around with unproven and unreliable technologies when your privacy is on the line. As such, we currently have no plans to introduce any features that rely on large language models or any other kind of generative AI.
If you are currently using an email provider that offers such features, we strongly encourage you to switch to a provider that doesn’t.
Photo by cottonbro studio.