Salesforce changes Slack API terms to block bulk data access for LLMs

Salesforce’s Slack platform has changed its API terms of service to stop organizations from using Large Language Models (LLMs) to ingest the platform’s data as part of its efforts to implement better enterprise data discovery and search.

While it sounds like a largely technical tweak, it could nevertheless have profound implications for a raft of internal and third-party AI apps that organizations have started using to tame data sprawl.

The new policy was outlined under the new heading Data usage in the latest terms of service, published on May 29. It prohibits the bulk export of Slack data via the API, and explicitly states that data accessed via Slack APIs can no longer be used to train LLMs. Instead, organizations will have to rely on the company’s new Real-Time Search API, which offers search only from within Slack itself.

It also restricts distribution of applications built using the API, requiring developers to either enter into a partner agreement with Slack or Salesforce, or only distribute their apps through the Slack Marketplace.

This could be inconvenient news on several fronts. First, third-party data discovery app makers will lose access to an important data source, with one, the makers of the Glean app, reported by The Information to have already emailed its customers to explain the move’s negative implications.

Second, Salesforce/Slack risks upsetting customers using their own LLMs. Slack is a hugely popular messaging system inside many organizations, and losing the ability to pull this data into larger internal discovery and search LLMs will be a loss.

In a blog expanding on the change, Slack justified the change as part of a broader overhaul of how AI and security intersect.

“With the Real-time Search API, approved partners and developers can retrieve exactly the data you need from Slack in real-time by searching messages and files directly, enabling safe, AI-powered use cases such as federated search and deep research,” the company said.

An alternative interpretation is that Salesforce wants to promote its own (or approved) AI tools available via the Slack Marketplace. Pessimists believe that this model could herald a wider clampdown intended to drive customers towards proprietary solutions.

SaaS overload

SaaS platforms – Salesforce, Teams, Gmail, Google Drive, Office 365, Dropbox, ServiceNow, GitHub, and many others – are hugely popular, but the more of these an organization uses, the more ‘silos’ of data employees must trawl through to find useful information.

Founded by former Google employees, Glean itself supports 100 of these platforms, which puts the scale of the challenge into perspective. Enterprise search LLMs solve this problem by ingesting data from multiple platforms through public APIs, allowing customers to query data through a single SaaS interface.

Meanwhile, AI capabilities have become critical to large platforms such as Slack, which might explain why Salesforce sees this capability as worth changing its API access rules to protect.

A Salesforce statement sent to this publication on Wednesday directed us to the earlier Slack blog post, saying that data security was at the heart of the change:

“A cornerstone of Slack’s new data connectivity strategy is enabling real-time search access via our Real-time Search API. This allows users to interact with data directly where it resides, without the need to duplicate or move data and permissions between systems,” said a spokesperson. Going forward, this would allow access on a rate-limited basis.

“This API also eliminates the need for large data exports from Slack, keeping customer data secure, while maintaining support for key use cases like permission-based search,” the spokesperson added.

However, this also rules out ingesting Slack data into an external LLM, a restriction that has been viewed with skepticism by industry sources.

“On the surface, this feels like Salesforce pulling up the ladder. For organizations trying to build internal copilots or unify siloed data across apps, this move kneecaps a lot of that progress,” commented Wyatt Mayham, CEO of Northwest AI Consulting.

“That said, I get the liability angle. Letting third party models train on private messages creates legal and privacy risks Salesforce doesn’t want to own. But if this is a step one toward a strategy where Slack data becomes a monetizable asset for licensing, that’s a different story,” he added.

“If this spreads, we’re looking at a fractured AI app landscape. The more platforms wall off user data under the banner of ‘protection’ the harder it gets to build anything that connects context across tools, which is what most teams actually want AI to do,” said Mayham.

AI expert and speaker on AI, Bob Hutchins of Human Voice Media, agreed on the wider implications.

“This move by Slack/Salesforce is part of a broader pattern; we’re seeing platforms tightening their grip on user data under the banner of security or product integrity, but often in ways that primarily serve their own AI ambitions,” said Hutchins.

“Let’s call it what it is – platform enclosure.” he said. “For organizations trying to reduce data silos, this creates friction. Many are turning to AI to surface insights across platforms — email, docs, project tools, messaging apps like Slack. Blocking third-party access could mean fewer choices, more workarounds, and slower decision-making. If this tactic spreads to other vendors, we’re talking about a future where your data becomes less yours and more theirs.”