FAQ
What is basebox?
Secure AI management system
Why do I need basebox in my company?
The controlled use of AI is a key challenge for companies. While the strategic potential is clear, there is often no secure way to integrate it into existing structures. basebox is the management system for the confident use of AI – with full control over data, processes and costs.
How does basebox work?
Control
The basebox AI Management System controls all AI operations within the enterprise.
Departments and teams receive defined access rights, usage scales according to
needs, and costs are allocated based on causation. All AI activities are transparently
documented.
Integration
Integration occurs through standardized interfaces into the existing IT landscape.
Employees use their familiar login credentials, departments continue using their
established systems. Data access is controlled and purpose-bound.
Security
Through on-premise or private cloud operations, sensitive data remains within the
company. Precise access control and a closed data circuit guarantee the highest
security standards while maintaining maximum flexibility. The system meets the strict
security requirements of German hospitals that are already using basebox.
What does basebox cost?
Free version (Free)
- 0€ per year
- For 2 users
- Standard functions
- 1 million cloud tokens
- Unlimited on-premise tokens
Small package
- 2.900€ per year
- For 10 users
- Standard features
- 12 million cloud tokens
- Unlimited on-premise tokens
Medium package
- 11.900€ per year
- For 50 users
- Standard features
- 60 million cloud tokens
- Unlimited on-premise tokens
Enterprise package
- 48.900€ per year
- For up to 300 users*
- Standard plus Enterprise features***
- 360 million cloud tokens
- Unlimited on-premise tokens
* Each additional user €108/year (both variants)
** Role Based Access Control (RBAC), RAG, Assistant, Model, Apps
*** OpenIDConnect, LDAP, SSO, REST-API
Who needs basebox?
All companies and organizations that want to become more efficient with AI and work with confidential data.
- Hospitals, clinics, medical practices
- Insurance companies, financial service providers, banks
- Courts, law firms
- public authorities
- Defense companies
- technology
- Research companies
- SMEs
Why can't I just use ChatGPT, Microsoft Copilot or similar providers?
- Because you give your data and your company knowledge to the outside world when your employees send a prompt to these providers.
- You also send your data to basebox – but nothing is saved.
- basebox is a European company and is subject to the relevant regulations.
- For companies with particularly high security requirements, basebox will provide an on-premise solution.
How does basebox ensure that the AI's answers are correct?
- basebox uses the best open source AI models.
- Various techniques are used to increase the accuracy and reliability of the answers.
- Despite all care, AI models are software programs.
- Incorrect statements are possible.
Can basebox prevent AI hallucinations?
Not prevent it, but reduce it to a minimum.
basebox uses various technologies for this purpose. For example:
- Strict rules on how the AI models may answer (e.g. that they may not make any assertions).
- Use of specialized AI for technical questions.
- The option of storing your own company data to which the answer may refer.
What risks do I have when using basebox?
There are no risk-free products. But basebox has security and data protection as its foundation.
Security measures to protect users’ data and privacy
- The communication between your device and basebox is end-to-end encrypted so that no one can read your data.
- basebox is a web application. There is no local installation.
- basebox does not store any data.
- The servers are operated in Europe.
- basebox has been tested by security experts who were unable to find any security gaps despite intensive attack attempts.
- The software is programmed in Rust, a language known for its high level of security.
The content-related risks of using AI are manifold. Hallucinations cannot (yet) be avoided. basebox is working on many individual technologies to reduce hallucinations.
How do I protect my IP?
basebox does not store any data. Neither the user’s requests nor the AI’s answers.
Your data belongs to you. When your employees send a prompt to basebox, the basebox AI has 100% dementia. This means that the AI processes your prompt, sends the answer back to you and forgets everything on the spot.
Can basebox also process images, audio and videos?
Currently, basebox is specialized in text processing.
Images, audio and video will be possible. This will make basebox even more versatile and suitable for more and more applications.
Does basebox have a RAG?
Yes, a very fast, high-quality RAG. You can create your own apps in basebox and process large amounts of data in them. This is particularly suitable for knowledge management. E.g. intranet information, Confluence connection, wikis, SOPs, guidelines and much more.
Do I need my own data to use basebox? If so, which ones? Why?
No, you do not need your own data. However, you can significantly improve the quality of the answers if you use your own data.
Does basebox use ChatGPT?
No.
- basebox uses open source AI models for on-premise hosting.
- For cloud use, basebox uses Claude from Anthropic. Instead of the Claude API, the model is used in Google Vertex Hosting. This means that no data is sent to Claude.
Which AI models does basebox use?
In each case, the best open source models for the on-premise version (this is the version when basebox is hosted on your own company servers).
With the on-premise model selection, it depends on the performance of the available hardware. A very suitable model is currently Llama 3.3 70B from Meta.
For individual apps, basebox uses models specialized for the respective task.
Can I train my own LLM?
No.
basebox is not intended for training your own models. basebox concentrates on the use and application of already trained models.
However, it is easily possible to integrate your own, already trained models into basebox. The transfer of models from platforms such as Hugging Face is then particularly easy.
Can I connect basebox to my existing systems (ERP, CRM, HIS, etc.)?
Yes, basebox has a REST API that can be used to connect any application simply and securely.
Where is basebox hosted?
Cloud version
- Hosting takes place on secure, GDPR-compliant servers in Europe.
- Our partner is the company Hetzner.
- The AI models are hosted in the Google Cloud.
- If the API regulations of an open source model manufacturer are in line with the basebox requirements for security and data protection (no data is stored), we also use these.
On-premise version
- The AI management system and the models are hosted on your own computers. In your data center, or your private cloud. You make the decision yourself.
How complex is it to get basebox running on your own servers?
Cloud version
- No separate servers are required to operate the basebox Cloud version.
On-premise version
- With the on-premise version, installation takes just a few minutes.
- All components for secure operation are included (models, security mechanisms, auth services, databases, RAG, REST-API, LDAP, RBAC) and can be installed and operated as a complete package.
Do I need my own server for hosting?
No, if you use the cloud version:
- You do not need your own server for hosting.
- Your basebox is created as a separate instance in the basebox cloud.
- basebox is operated via the browser and runs smoothly on a laptop.
Yes, if you use the on-premise version:
- Installation is very simple.
- All components are included.
- Especially the AI models. If you do not want to use the models supplied, you can use your own models. The connection is made via API.
- We will give you recommendations for the server configuration.
Is basebox open source?
No.
Is basebox a medical service?
No.
- basebox is an AI management system that enables the secure use of AI applications. Even though basebox is currently mainly used in the healthcare sector, it is suitable for all industries.
