AI companies have always been regarded as shifty when it comes to the training data they use, particularly where they get them. It's time for people to read the fine print as they might unknowingly agree to have their data used to train AI. Such is the situation with Slack.

Slack Uses Customer Data to Train AI

Machine learning AI models need troves of data to become smarter, therefore providing its users with better service. However, AI companies haven't always been upfront about the kind of data they use. OpenAI has been under fire for it, and now Slack is facing the same backlash.

DuckBill Group Executive Corey Quinn decided to take a look at Slack's Privacy Principles, only to find that the messaging app has been using user data to train AI and machine learning models. The problem is that customers have already agreed to it probably without their knowledge.

"To develop AI/ML models, our systems analyze Customer Data (e.g. messages, content, and files) submitted to Slack as well as Other Information (including usage information) as defined in our Privacy Policy and in your customer agreements," the policy reads.

Since it's an opt-out feature, the app already takes and uses the data by default, as reported by Engadget. Slack is not making it easy to opt out of the policy as well. To do so, customers have to email the company instead of having a toggle button that turns it off.

What's worse is that not everyone is eligible to send that opt-out email. The privacy notice states that only the organization, workspace owners, and primary owners can contact customer support to stop the leeching of data.

Slack clarified that the platform has "machine-learning models for things like channel and emoji recommendations and search results," and that the machine-learning models in question are not generative. That means that your private data will not pop up when a person enters a certain prompt.

Even Slack's policies are contradictory. In another published policy, the messaging app states that when developing AI or machine-learning models, "Slack can't access the underlying content," and that they have "various technical measures preventing this from occurring."

The privacy policy statement is the direct opposite of another, which assures customers that they can work without worry as their data is their data. "We don't use it to train Slack AI," the company notes. "Everything runs on Slack's secure infrastructure."

Read Also: Salesforce to Pay a Whopping $27.7 Billion to Acquire Slack

What Slack Uses It For

Since it gets confusing how Slack uses customer data in the first place, here are the listed improvements that the messaging platform claims to develop as it uses private data such as messages, content, and files. 

One is Channel Recommendations, wherein Slack uses insights to recommend that a user join a new public channel in their company. Slack will scan the membership, topics, and activity overlaps to see where a user may fit in.

There's also Search Results, where the company bases the outcomes on the customer's historical search results and previous engagements. Autocomplete, provides suggestions to complete search queries by sourcing common public message phrases in a workspace.

Related: Slack Debuts AI Tool to Summarize Long Messages, Replies