Protects against more than 95% of aggressive texts.
Respond within milliseconds. 24/7.
Gain time for what matters.
Our systems constantly adapt to change.
Trained on the basis of various discourses.
18 pretrained, many more supported. See the Docs.
Unbiased and socially fair. No stereotypes. See the Model Cards.
Thematically neutral. Controversial topics welcome.
Efficient and resource-saving architecture.
All AI models are accessible via an Application Programming Interface (API). The API is to be used programmatically, i.e. through the use of a programming language and one of its standard libraries.
The API consists of a REST endpoint secured by SSL. Authentication happens via a JWT bearer token that can be regenerated at any time in the account page. Data submitted for classification over this encrypted channel is directly passed to the classification model which returns a simple number, its score.
See the Documentation for more information.
Due to its simple design and standardized interface, the API can easily be integrated into any existing system. The only preprequisite is access to the system's source code.
Absolutely not, you may cancel anytime.
Data you send to the API never gets saved or copied. Our services operate from self-managed infrastructure within Austria. There are no dependencies on third-party services. All data are protected by the GDPR and Strong Customer Protection directives of the European Union.
We are contributing to current research in the area of dynamic neural training. Our focus is in de-biasing neural classification models using methods to increase information flow between layers of the neural architecture and thus avoiding the general problem of overfitting for words which are prominent in the training data. This allows us to provide generally de-biased classification regarding identity terms, controversial topics, as well as spurious correlations a neural architecture might pick up on.