AI Rules: Difference between revisions
Jump to navigation
Jump to search
No edit summary |
|||
| Line 13: | Line 13: | ||
To protect your data, the following measures are in place: | To protect your data, the following measures are in place: | ||
* Chat messages are stored within your [[Project]] and follow the same data and privacy terms described in your license agreement. | * Chat messages are stored within your [[Project]] and follow the same data and privacy terms described in your license agreement and our [[ISO_27001]] certification. | ||
* All models run exclusively on [[Cloud Computing|Tygron-owned GPU infrastructure]] in the Netherlands. | * All models run exclusively on [[Cloud Computing|Tygron-owned GPU infrastructure]] in the Netherlands. | ||
Revision as of 12:02, 8 October 2026
The Tygron Platform uses AI Agents in several features, most prominently the Chat Channels. In order to properly use this technology we have three rules build into the design and usage:
1 Safety
For safe usage several measures have been implemented:
- Scope limitation: The AI can only be used for Tygron Platform related question & tasks. This is also where it has been tested for. It should not be used to answer question outside this scope.
- The AI Agent can also directly interact with your project and edit data using tools. For example creating a new dashboard, indicator or changing an overlay. However AI can make mistakes and if you do not want this behavior you can disable "AI Editing" via the Chat panel or setting Setting AI_EDITING to "false".
- Not all API endpoints can be accessed by the AI. For example root API endpoints are restricted, this prevents the AI from saving or deleting projects. Furthermore the AI cannot create/start additional AI Agents or change the previously mentioned AI_EDITING Setting to "true".
2 Privacy
To protect your data, the following measures are in place:
- Chat messages are stored within your Project and follow the same data and privacy terms described in your license agreement and our ISO_27001 certification.
- All models run exclusively on Tygron-owned GPU infrastructure in the Netherlands.
- Zero Data Retention (ZDR): no data or metadata is stored by the AI model and conversations are not used for AI training.
3 Efficiency
To make the AI usage more energy efficient:
- All Generative AI (GenAI) models used are small and optimized for the Tygron Platform. This means that in comparison with general purpose frontier models they require much less memory and compute power to reach the same results within the Tygron Platform scope. Less memory and compute power also means a smaller GPU infrastructure and lower power usage (KwH).
- Transparent power usage: Via the logs API endpoint it is now also possible to view the power usage (in KwH) per simulation job type during the day. This gives insight into the power usage of Overlay simulations and AI Agents.
- The Tygron-owned GPU infrastructure is optimized for energy efficiency. For example when there are no active simulation jobs the GPU hardware is powered down in a sleeping mode.