OpenAI will soon launch a beta test of a feature called ‘ChatGPT Connectors’ that will let businesses link ChatGPT to Slack and Google Drive.
The new feature was revealed in a recent document. It aims to help companies use existing files and conversations from those platforms to inform ChatGPT’s responses.
The ‘ChatGPT Connectors’ feature is set to roll out for ChatGPT Team subscribers first. These subscribers will be able to sync their Slack channels and Google Drive files so ChatGPT can reference that material when answering queries. The document indicates that ChatGPT Connectors will expand to other platforms in the future, including Microsoft SharePoint and Box.
According to the document, ChatGPT Connectors is powered by a specialized version of OpenAI’s GPT-4o model. The system can refine responses based on internal business data pulled from an encrypted copy of company files and messages stored on ChatGPT’s servers. The document explains that any extra information not used in an answer can be accessed by clicking on a sources button within ChatGPT.
The new OpenAI feature will fully respect Drive and Slack permissions.
OpenAI emphasizes permissions. According to the document, ChatGPT will not leak any customers’ private data. It further mentions that Google Drive and Slack permission will be “fully respected” and stay “continuously up to date.” If a staff member does not have access to certain items on Slack or Google Drive, those documents will remain hidden from ChatGPT.
Administrators can also decide which files and channels get synced. One downside is that employees might receive “substantially different” answers to the same prompt, depending on who has access to the relevant data.
There are some limitations to the first version of ChatGPT Connectors. Images in Google Docs, Google Slides, PDFs, and other supported file types are off-limits to the new feature, which can only read text data from those sources.
The document also states ChatGPT Connector is unable to retrieve direct messages or group messages on Slack, and it ignores any input from Slack bots. Sheets and Excel analysis is restricted to reading data, rather than performing calculations on those files.
OpenAI is asking companies that want to join the beta to provide a sample of 100 internal documents, spreadsheets, presentations, or Slack conversations for testing.
Although the company says it will not directly train its model with that content, the document suggests data could be used for “input to synthetic data generation.”
Still, OpenAI reiterates, “No data synced from Google Drive or Slack will be used for training.”
Cryptopolitan Academy: Tired of market swings? Learn how DeFi can help you build steady passive income. Register Now