OpenAI is testing new tools for ChatGPT’s deep learning feature, including connections to Slack, Canva, and Gmail, allowing for deeper analysis of information from work environments.
OpenAI is testing new tools for ChatGPT’s deep learning feature, including connections to Slack, Canva, and Gmail, allowing for deeper analysis of information from work environments.
OpenAI is adding support for new integrations (called connectors) for its «deep research» feature in ChatGPT, which automates online research. As BleepingComputer reports, Slack is among the first new connections.
The «deep research» function allows the user to ask a short query, and the system independently processes Internet sources to create a full-fledged research or analytical text. Thanks to new connectors, ChatGPT will be able to receive additional data from corporate environments, such as messengers or design services.
The «deep research» feature was introduced in the paid versions of ChatGPT in 2024 and allows you to create deep analytical texts on demand. It combines the capabilities of generative AI with search, fact checking and document structure building. OpenAI is gradually expanding its functionality, focusing on the needs of researchers, business analysts and journalists.
The Slack connection allows you to use the history of work chats to form the context of the research, for example, when preparing reports or technical documentation. However, it is currently unknown whether the integration is through official cooperation with Slack or through open APIs, which have a number of limitations.
Canva also supports «deep research.» This means that ChatGPT can generate not only text documents, but also visual materials using the given content.
OpenAI is also testing integrations for ChatGPT Search, including with Gmail, which allows for searching and summarizing conversations, opening up new scenarios for business users.
OpenAI assures that users' private data is not used to train models if the appropriate privacy setting is activated in the settings.
We previously reported that research shows that one in three links generated by language models like GPT-4.1 do not belong to the brand in question. Some of these addresses turn out to be dangerous, opening the way for phishing attacks and the spread of malicious code.
