OpenAI is gearing up to beta test “ChatGPT Connectors,” a new feature that will enable ChatGPT Team subscribers to integrate their workspace tools like Google Drive and Slack with the AI chatbot, according to documents recently reviewed by TechCrunch.
The innovative feature aims to expand ChatGPT’s capabilities by allowing it to access and analyze information stored across company accounts. With these connections in place, ChatGPT will be able to provide answers informed by files, presentations, spreadsheets, and conversations from connected platforms.
“This will allow employees using ChatGPT to easily make use of internal information similar to how they can use world knowledge via web search,” states the document outlining the initiative.
Expanding Business Integration
ChatGPT Connectors represents OpenAI’s latest strategic move to cement ChatGPT as an essential component of corporate software ecosystems. While the initial rollout focuses on Google Drive and Slack integration, the company plans to extend support to additional platforms like Microsoft SharePoint and Box in the future.
This development comes at a time when businesses show mixed reactions to AI tools. Some organizations remain cautious about allowing AI systems like ChatGPT access to sensitive internal information, while others have eagerly adopted such technologies across their operations.
The new feature could potentially sway hesitant executives to reconsider their stance, particularly as it promises to maintain strict security controls. Additionally, it positions OpenAI to compete directly with specialized AI-powered enterprise search platforms such as Glean.
Technical Capabilities and Limitations
Behind the scenes, ChatGPT Connectors is powered by a customized version of OpenAI’s GPT-4o model specifically designed to refine responses based on internal company knowledge. When the feature launches in beta, all users within participating ChatGPT Team workspaces will gain access to this specialized model through OpenAI’s ChatGPT apps.
The system works by creating a search index from encrypted copies of company files and conversations stored on ChatGPT’s servers. When queries are submitted, the model searches and analyzes potentially relevant internal information before formulating a response.
“Additional related information which the model did not directly make use of is accessible by clicking on the sources button at the bottom of each response,” explains the document. “When appropriate, the model will directly respond with a list of relevant results.”
However, several technical limitations exist in the current implementation. According to the document, ChatGPT Connectors cannot process images within Google Drive files (including Google Docs, Slides, PDFs, and various Microsoft Office formats). The system can only “read” but not analyze data in spreadsheets, and it cannot retrieve direct messages or group conversations from Slack. Messages from Slack bots are also excluded from processing.
Privacy and Permission Controls
Addressing potential privacy concerns, OpenAI emphasizes that Slack and Google Drive permissions are “fully respected” and “kept continuously up to date” within the system. This means ChatGPT Connectors will sync Slack private channel memberships, Drive file permissions, and directory information.
The document assures that employees won’t be able to discover content through ChatGPT that they couldn’t access directly in Google Drive or Slack. Additionally, administrators will have control over which Slack channels and Google Drive files are synchronized with the system.
One noted consequence of this permission-based approach is that employees might receive “substantially different” responses to identical ChatGPT prompts based on their access levels within the organization.
Companies interested in participating in the ChatGPT Connector beta program are being asked to provide OpenAI with approximately 100 documents, spreadsheets, presentations, or Slack channel conversations for testing purposes.
OpenAI states in the document that it won’t directly train its models on this information but may use the materials “as input to synthetic data generation” that could potentially be incorporated into training. The document explicitly states, “No data synced from Google Drive or Slack will be used for training.”
This careful positioning around data usage likely reflects OpenAI’s awareness of ongoing concerns about AI systems potentially learning from or leaking sensitive business information.
As of the time of reporting, OpenAI has not responded to multiple requests for comment regarding ChatGPT Connectors.
The development of these integration tools highlights the growing importance of context-aware AI assistants in the enterprise space, as companies seek to leverage internal knowledge bases alongside the broader capabilities of large language models.
