A mysterious privacy page that describes how users’ Slack conversations, including direct messages, are used to train what the Salesforce-owned company refers to as “Machine Learning” (ML) and “Artificial Intelligence” (AI) systems has alarmed Slack users on Mastodon, Threads, and Hackernews. The admin of your company’s Slack configuration must email Slack to request that these features be disabled in order to opt out of them.
According to the policy, Slack systems “analyse 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 agreement.” This policy is applicable to all Slack instances, not just those that have chosen to use the Slack AI add-on.
In other words, Slack uses everything you type to train its systems. Slack claims that there are “technical controls in place to prevent access” and that data “will not leak across workspaces.” Nevertheless, it’s common knowledge that discussions having AI chatbots are not private, thus it’s easy to picture things going wrong in some way. The business has to be providing something really alluring in exchange for the risk, right?
What advantages come with allowing Slack to train its AI on your data?
The section discussing the possible advantages of Slack incorporating all of your chats into a sizable language model claims that doing so will enable the company to offer enhanced channel recommendations, better autocomplete suggestions, better search results, and—I wish I was kidding—better emoji suggestions. Great if all of this seems helpful to you! Personally, I don’t think any of these features—aside from perhaps improved search—will significantly increase Slack’s use for completing tasks.
The emoji situation in particular is ridiculous. Slack is essentially stating that in order to improve its emoji suggestions, it must feed your conversations into an artificial intelligence system. Think about this real quote—I swear it’s from Slack’s website, not The Onion—that follows:
Emoji reactions to messages could be suggested by Slack based on the message’s sentiment and content, previous emoji usage, and how frequently the team uses the emoji in different situations. For example, we will recommend that users respond with 🎉 to new messages that are equally favourable if 🎉 is a typical response to joyful messages in a specific channel.
I am in complete amazement at the potential applications of this amazing technology and have completely given up worrying about privacy issues. AI really is the communication of the future.
How to choose not to participate in Slack’s AI training
The bad news is that Slack uses your conversation data to train its huge language model and you, as an individual user, cannot opt out of this. That can only be completed by a Slack administrator, which is typically someone from your company’s IT department. Moreover, administrators must email requesting that users opt out rather than relying on a button in the settings.
Slack’s actual words on the subject are as follows:
You have the option to opt out of having your Customer Data included in Slack global models. Please have your organisation, workspace owners, or primary owner contact our Customer Experience team at feedback@slack.com with the subject line “Slack global model opt-out request” along with the URL of your workspace or organisation in order to opt out. We will handle your request and get back to you as soon as the opt-out is finished.
This seems like a dark pattern where you make something unpleasant to do to get people not to do it. In response to the current outcry from customers, perhaps the corporation will simplify the opt-out procedure.
An alert that direct messages in Slack aren’t private
When I think about it, I find it kind of funny that my Slack data might help my previous employers with their search and emoji recommendations. During my former employment, I often texted coworkers negative messages about my manager and the company’s leadership. Every time a specific CEO is named, I can almost see Slack suggesting a certain emoji.
Though it’s a funny thought, the entire event should serve as a helpful reminder to workers worldwide that their Slack direct messages aren’t truly private. On Slack, nothing you say is private, not even in direct messages. While Slack does utilise such data to train bots like these, your employer can also simply view those private communications. If you must trash speak your firm, I strongly advise utilising anything unrelated to them. Could I make the suggestion that Signal?