(Credit: Slack/Salesforce)
Slack is training its machine learning features on its users' data—and everyone's opted-in by default.
Slack uses machine learning, a subfield of AI, to operate in-app features like channel recommendations, search results, autocomplete, and emoji suggestions, according to the company. But it's pulling from users' Slack messages, data, and files to improve these features.
Opting customers into something like AI model training by default may be frustrating for those who would have preferred advance notice. When Slack began using its customers' messages, many users were unaware of this change and may have wanted to have been directly notified and given the chance to opt-out before the data collection and use began in the first place.
Corey Quinn, an executive at cloud computing billing company DuckBill Group, pointed out Slack's policy on AI late Thursday night in a frustrated post. "I'm sorry Slack, you're doing fucking WHAT with user DMs, messages, files, etc?" Quinn wrote.
Slack replied to Quinn's post by reiterating its policies which are that yes, the Salesforce-owned company is using Slack users' content to train some of its AI-powered app tools. It confirmed that it's not opting users or companies out by default, and it's not using customer data to train its paid generative AI tool. It also said that companies or workspace administrators (not individual users) are able to request via email to have their messages and content opted-out of the training dataset.
On X, over half a dozen users argued that Slack should have opted users out by default and taken an opt-in approach instead. Lawyer and Silver Key Strategies consultancy founder Elizabeth Wharton called Slack's decision a "privacy mess" and criticized the company for not giving individual users an easier way to opt-out.
PCMag reached out to Slack for comment on when exactly this AI privacy policy went into effect, why Slack has chosen to take an opted-in default policy, and what external vendors it's working with on its machine learning tools. A Slack representative replied with a statement that reiterates what it previously wrote on X.
"We do not build or train these models in such a way that they could learn, memorize, or be able to reproduce some part of customer data," the representative said. "Customers can exclude their data from helping train those (non-generative) ML models."
Slack's policies may confuse those who see its AI page, which proclaims: "Work without worry. Your data is your data. We don't use it to train Slack AI." While Slack says its paid generative AI tools aren't being trained on Slack user data, it is training some of its other machine learning models with user data. So the implication that user data fully belongs to users (and isn't being used by Slack) is broadly untrue.
Slack isn't the only tech platform that's not opting users out of AI tool training by default. The website hosting and design platform Squarespace quietly rolled out a feature last year allowing its users to opt-out of some AI crawling tools that scrape web data—but AI crawling is permitted by default. When using Squarespace, you have to manually navigate through the platform's settings to find the toggle to opt-out of AI crawlers (go to Settings > Website > Crawlers > AI Crawlers). Squarespace doesn't guarantee this will fully stop AI tools from using your site's data, however.