safeGPT
our proof of concept for a secure chatbot.
safeGPT is our proof of concept for a secure chatbot. With safeGPT, you can utilize the power of cloud-based LLMs like ChatGPT while minimizing the risk of disclosing sensitive data.
Increase your productivity while maintaining a high standard of data security!
ChatGPT has taken the world by storm. Many companies use the chatbot to increase their productivity. But anyone who “talks” to ChatGPT runs the risk of revealing things that are no one else’s business – personal or confidential information, for example, customer data or company-internal expertise.
Our proof of concept safeGPT helps you to utilize ChatGPT securely. safeGPT is based on a two-step filter that removes as much sensitive data as possible from the prompts, such as names, addresses, amounts of money, etc. This filter runs on your local infrastructure so that the sensitive data does not leave your company. The anonymized requests are then sent to a powerful cloud-based language model. safeGPT stores the sensitive data locally and later reinserts it into the responses before they are transmitted to the user. Ideally, the users do not realize they are talking to ChatGPT through a local filter. The user experience does not suffer, and at the same time a high level of data security is ensured.
Try out the functions of safeGPT.
Please note:
- The current version of safeGPT is a proof of concept and not a final product.
- Currently, safeGPT is running on an external infrastructure, of course, but could also run locally at your premises tomorrow.
- We do not store any of your prompts. Nevertheless, we ask you not to enter any particularly sensitive data. Use fictitious data like „John Doe” and test safeGPT with it.
test_safegpt
We support with strategic consultation, drive the implementation of Generative AI use cases and enable the organization.
Katharina Fulterer
Partner
Financial Services
Generative AI | Data & AI
Lukas Fischer
Senior Consultant
Public Services
Data & AI