Salesforce just changed its Slack API! Now blocking bulk data access for LLMs. What does this mean for your AI apps? #Salesforce #SlackAPI #LLMs
Explanation in video
Big News from Slack: How They’re Changing the Way AI Sees Your Messages!
Hey everyone, John here! If you’re using the internet, you’ve probably heard a lot about Artificial Intelligence, or AI, lately. It’s that super-smart computer technology that’s learning to do all sorts of amazing things. Today, we’ve got some interesting news from a company many of you might use for work or in your communities: Slack. They’ve just made a change that affects how AI can use the information shared on their platform. Sounds a bit complicated, right? Don’t worry, we’re going to break it down together!
First Off, What’s Slack and Why Is This a Big Deal?
Imagine a giant online chatroom for your company or group. That’s kind of what Slack is! It’s a place where teams can send messages, share files, and keep track of projects. It’s super popular, and because so much information flows through it, some companies thought, “Hey, what if we could use AI to help us understand all this data?”
For example, an AI could potentially help find old conversations quickly, summarize long discussions, or even spot trends in what people are talking about. This could be really useful for businesses trying to stay organized and efficient.
Lila: “Hi John! So, Slack is like a big digital office where everyone talks? And companies want AI to be like a super-helpful assistant that reads everything?”
John: “Exactly, Lila! You’ve got it. And that’s where this new change comes in. Salesforce, the big company that owns Slack, has decided to adjust the rules about how AI can ‘read’ all that Slack information.”
So, What Exactly Did Slack Change? The Nitty-Gritty, Made Simple!
Okay, here’s the main thing: Slack used to have a way for other computer programs (including AI) to ask for and get large amounts of data from Slack. Think of it like being able to download a huge chunk of all the conversations and files.
Lila: “John, hold on a sec. You said programs could ‘ask for’ data. How does a computer program ask another computer program for something? It sounds like a secret handshake!”
John: “Haha, that’s a great question, Lila! It’s not quite a secret handshake, but it’s close. They use something called an API. API stands for Application Programming Interface. Think of an API like a waiter in a restaurant. You (a program) tell the waiter (the API) what you want from the kitchen (Slack’s data). The waiter then goes to the kitchen, gets your order, and brings it back to you. So, these AI programs were using Slack’s API to order up big batches of conversation data.”
Now, Slack has changed the “menu,” so to speak. The new rules, which came out at the end of May, basically say you can’t use these APIs to do a “bulk export” – that means taking out huge, massive amounts of Slack data all at once, especially if the plan is to use that data to train certain kinds of AI.
Lila: “Okay, API as a waiter makes sense! But now you’ve said they can’t use the data to train ‘certain kinds of AI.’ You mean those LLMs I keep hearing about?”
John: “You’re spot on, Lila! LLM stands for Large Language Model. These are the super-smart AIs, like ChatGPT or others, that are amazing at understanding and generating human-like text. To get that smart, they need to ‘read’ and learn from enormous amounts of text data. Some companies were hoping to feed their Slack conversations to these LLMs to make them better at understanding their specific business chatter or to build custom AI tools.”
So, the new rule is: data you get from Slack’s APIs can no longer be used to train these LLMs. It’s a pretty big shift!
Why Did Slack Make This Change? What’s Their Goal?
This might seem like a technical detail, but it could have some pretty big effects. The original article mentions that companies have been using AI apps (both ones they build themselves and ones they buy from other companies) to help them “tame data sprawl.”
Lila: “John, you mentioned ‘data sprawl.’ Is that like when my desk gets super messy with papers everywhere, and I can’t find anything?”
John: “That’s a perfect way to think about it, Lila! In a company, ‘data sprawl’ means there’s information scattered all over the place – in emails, in documents, in chat messages like those on Slack. It gets overwhelming and hard to find what you need. AI was seen as a way to help sort through all that mess, even the mess within Slack.”
So, why the change? Here are a few possible reasons:
- Keeping Data Within Slack: Salesforce and Slack likely want to encourage companies to use Slack’s own tools for searching and understanding their data. They’ve introduced something called the “Real-Time Search API.”
- Control and Security: Allowing massive exports of company conversations could have privacy or security risks if not handled perfectly. This move gives Slack more control over how its data is accessed and used.
- New Product Strategy: Slack might be planning to offer its own advanced AI features directly within Slack. If everyone is building their own tools with Slack data, it might compete with what Slack wants to offer.
Lila: “So, if companies can’t easily take all their chat data out to train their big AI helpers, how are they supposed to find information or make sense of it all within Slack? You mentioned a ‘Real-Time Search API’?”
John: “Exactly! Slack isn’t leaving companies completely in the dark. This new Real-Time Search API is designed to let companies search their Slack data, but it seems to be focused on searching from within Slack itself, rather than exporting it all for external AI training. Think of it like Slack saying, ‘Instead of taking the whole library home, use our excellent library catalog and reading rooms right here!'”
What Does This Mean for Businesses Using Slack and AI?
This change could definitely shake things up for some businesses:
- Internal AI Projects: Companies that were building their own custom LLMs or AI tools trained on their internal Slack conversations might need to rethink their approach. They won’t be able to easily get that bulk data anymore for training.
- Third-Party AI Apps: There are also many other companies that make AI tools designed to plug into services like Slack. If these tools relied on bulk data export and LLM training, they might need to adapt or they might not work with Slack in the same way.
- Focus on Slack’s Tools: Businesses will likely need to rely more on Slack’s built-in search capabilities and this new Real-Time Search API. This could be good if Slack’s tools are powerful, but it might be limiting if companies had very specific, custom AI needs.
- Data Discovery Shift: The way companies “discover” or find information within their vast Slack archives will change. Instead of training an external AI brain, they’ll be using Slack’s internal search mechanisms more.
It’s a bit like if you used a special app to analyze all your emails to find patterns, and then your email provider said, “Sorry, you can’t download all your emails in bulk for that app anymore, but we have a new, powerful search feature you can use instead!” It changes how you interact with your own information.
Our Quick Thoughts on This
John’s Take: From my perspective, this seems like a strategic move by Salesforce. They’re likely looking to create a more controlled ecosystem around Slack data. It could be about ensuring data privacy and security, but it’s also very likely about guiding users towards Slack’s own upcoming AI-powered features. If Slack can provide powerful, built-in AI tools, they become even more indispensable to businesses. It’s a common strategy: offer a platform, then offer premium tools to make the most of that platform.
Lila’s Take: As someone still learning all this, it makes me think… On one hand, I get why a company would want to keep all its chat information more private and secure. It feels a bit like if I had a diary, I wouldn’t want just anyone to be able to copy all of it easily. But on the other hand, if AI could help me find important notes in my super-long diary quickly, that would be amazing! So, I guess it’s a balance. It’s interesting that Slack is saying, “Use *our* special magnifying glass, not one you built yourself or got from somewhere else.”
Wrapping Up
So, there you have it! Slack is changing how companies can access and use their Slack data, especially when it comes to training those big AI models called LLMs. They’re steering users towards their own internal search tools. This will likely mean some adjustments for businesses that were relying on broader data access for their AI projects.
It’s another example of how companies are figuring out the rules of the road in this fast-moving world of AI. We’ll be keeping an eye on how this develops!
This article is based on the following original source, summarized from the author’s perspective:
Salesforce changes Slack API terms to block bulk data access
for LLMs