For amberSearch, the protection of data is the most important aspect
That is why we have set data protection and data security as our top priority in the development of amberAI from the very beginning.
This results in the following manifested principles:
- amberAI never queries external service providers or APIs. All models are run directly by amberSearch on appropriate servers.
- All development and processing steps take place on servers in Germany which are exclusively operated by German companies. Thus, no access to this data by US government agencies is possible. All servers and data centers are BSI C5, ISO 27017 and SOC 3 certified.
- amberSearch never has access to specific user questions or answers. (In contrast to ChatGPT and OpenAI, which grant themselves the right to analyze them manually and also publish them).
- Interactions are not used to improve AI models without explicit user consent. This is only possible upon request and consent.
- amberAI avoids, as far as technically possible, answering clearly racist, sexist or offensive questions.
The technical background
How does amberAI manage to deliver intelligent and appropriate answers and at the same time fulfill the high self-imposed goals of data protection?
To achieve this, we spent more than half of the development time exclusively on these points and incorporated them into our concept from the very beginning.
amberAI is based on self-adapted models, prompts adapted to them and individual contexts. These AI models can thus be executed completely autonomously on correspondingly powerful servers (also offline).
Furthermore, due to the currently exponentially increasing quality of ever new AI model architectures, it is not necessary to adapt the models individually to customers or to retrain them. In this case, the new architectures that are published in the meantime would often already be outdated before the finetuning can be completed accordingly.
Read more about this on our blog: https://ambersearch.de/de/private-ki/