About

Last updated: January 12, 2024

EgmontGPT, a private chat UI designed specifically for Large Language Model (LLM) interactions. EgmontGPT provides a secure and confidential way to communicate with LLM's, allowing you to leverage the power of artificial intelligence without compromising the privacy of your organization or leaking confidential data. You can safely use EgmontGPT for copy-righted text, business sensitive data or similar. Additionally, EgmontGPT is built to be vendor-agnostic, meaning you can use different LLMs from different providers including Open Source models. This allows us to choose the right model for our needs, such as model capabilities or cost-effective.

Privacy

All chat messages sent through EgmontGPT are hosted in a secure database for the purpose of providing chat history. They are not accessed by anyone other than yourself, and are deleted if deleted in EgmontGPT. No data sent is used for model training or other such reasons.

GPDR data

EgmontGPT is not intended for use with personal data as defined by the General Data Protection Regulation (GDPR). This includes both sensitive and non-sensitive personal information. You must anonymize any personal data before using EgmontGPT.

Because the chat history content is not monitored, we do not classify the data stored and therefore cannot comply with GDPR regulation, such as deleting personal data upon request (Right to be forgotten).

This does not mean EgmontGPT is insecure, but that it is not designed for personal data.

Data retention

Individual chat conversations are stored for 30 days after last update, after which they are automatically deleted.

When you enable web search, the data you send in the chat - while web search is enabled - is no longer secure. It is sent through a search engine such as google and hence may be stored through these services. Therefore, do not send any confidential information in the chat, when web search is enabled. Start a new chat when you wish to utilize web search to ensure no data leaked.

Technical details

  • EgmontGPT has been developed and is maintained by the Data & Intelligent Tech team in Egmont IT.
  • Users of EgmontGPT are authenticated through their Egmont user account.
  • The chat history data is hosted on an encrypted database, with privileged access management. No data is accessible outside of Egmont.
  • GPT 3.5 and 4 LLM backends are provided by Azure OpenAI Services deployed to a private Egmont Azure Subscription. The instances are exemption from abuse monitoring and human review. The models are stateless: no prompts or generations are stored in the model. Additionally, prompts and generations are not used to train, retrain, or improve the base models. For more details about the Azure OpenAI Service data privacy see: https://learn.microsoft.com/en-us/legal/cognitive-services/openai/data-privacy.

How to

  • To begin using the chat model, simply input your question in the chatbox on the frontpage. The model will respond, create a chat with a headline and you can continue the chat.
  • To start a new chat simply click 'New Chat' in the upper left corner. The EgmontGPT chat has a light mode and a dark mode. You can find this under 'Theme' in the info pane in the bottom left corner. Here you also find Q&A which answers the most asked questions, a Roadmap which contains the currently planned development features as well as a feedback form for you to request any features or improvements, and Settings where you can delete your entire conversation history. On the frontpage you will find a 'NEW' banner, highlighting the new changes to EgmontGPT. Furthermore, you will find a model card, showing you the current model you are using. If you click the arrow on the right side of the card, you can choose between available models and also alter the system prompt for that model if you wish.

Changelog

1.4.3 (2024-09-16)

  • Changed Roadmap to reflect new changes and previous
  • Changed Default model to GPT-4o

1.4.2 (2024-06-12)

  • Added New GPT 4 Omni model

1.4.1 (2024-06-12)

  • Changed Roadmap

1.4.0 (2024-04-12)

  • Added New Cohere model
  • Added Feature to change model parameters in assistants

1.3.3 (2024-04-09)

  • Added Model in datadog usage
  • Added Newest HF changes, including sort possibility

1.3.2 (2024-04-03)

  • Fixed Bug fixes for external logons

1.3.1 (2024-03-25)

  • Fixed Pagination on assistants

1.3.0 (2024-03-20)

  • Changed Updates to assistants feature
  • Added Added new models feature

1.2.0 (2024-03-01)

  • Changed Added new assistants feature

1.1.2 (2024-01-12)

  • Changed Updated Privacy section of About page
  • Changed Backend infrastructure improvements

1.1.1 (2024-01-02)

  • Added Link to guides & help in sidebar
  • Removed Christmas theme

1.1.0 (2023-12-22)

  • Added Web search feature
  • Added Small UI updates including chat sorting and new settings menu
  • Added New gpt-3.5-turbo model

1.0.0 (2023-11-23)

  • Added Separate Q&A section & Roadmap section
  • Added Feedback form
  • Added New gpt-4-turbo model preview
  • Added Christmas theme
  • Changed Minor UI improvements, including emojis, emojis everywhere
  • Changed Backend improvements

0.5.1 (2023-10-09)

  • Added Q&A section to About page

0.5.0 (2023-09-11)

  • Change Updated to OpenAI ChatGPT 4
  • Added New Egmont color scheme

0.3.0 (2023-06-15)

  • Changed Switched to React frontend
  • Changed Split frontend and backend API deployments
  • Removed Google Palm 2 / Bard model for security reasons

0.2.0 (2023-06-12)

  • Added Google Palm 2 model
  • Added App version to sidebar

0.1.1 (2023-06-12)

  • Added Egmont Chat title
  • Changed Support info

0.1.0 (2023-06-12)

  • First POC release built on Streamlit
  • Added ChatGPT 3.5