Claude Tag: AI arrives in Slack with immense power

Since GitHub Copilot in the code editor and Microsoft Copilot in Teams, the enterprise productivity sector has been methodically exploring how to integrate AI directly into existing work tools rather than asking teams to open a new window.Claude Tag builds on this principle and applies it to Slack, with one notable feature: unlike traditional chatbots that you query from time to time, Claude Tag can run continuously in the background, in what the company calls “ambient” mode.

Concretely, what does Claude see in your channels?

An administrator adds Claude Tag to a Slack workspace, grants it access to certain channels, and connects it to the approved tools and data sources. Then, all you have to do is mention @Claude in a conversation to delegate a task to it: it breaks the task down into steps, works asynchronously, and posts its results in the thread. The entire channel can see what Claude is doing, making it a shared resource rather than an individual assistant, since several members of a team can pick up a task started by a colleague, including the steps handled by the AI. Claude Tag runs on Claude Opus 4.8, Anthropic’s latest model, and replaces the former Claude in Slack integration (automatic migration scheduled for August 3 for those who do not opt in beforehand).

To distinguish Claude Tag from the many bots (with or without AI) already found on Slack, Anthropic highlights its “ambient” mode. When this option is enabled, Claude Tag scans the channels it has been authorized to access, proactively identifies information it considers relevant, and sends alerts or summaries.Anthropic recommends not enabling this mode immediately, while teams come to understand the system’s “failure modes” (a phrasing that has the merit of being straightforward). Administrators retain control over access rights, available tools, and spending limits per channel and per organization. Activity logs make it possible to verify what Claude did and in response to which request.

An AI that reads everything continuously: what are the risks to your data?

Having an AI continuously read the conversations in a company Slack channel is no small matter, even if platforms of this kind have accommodated it since the early days of workflow automation. The difference here is the extent of the context that Claude Tag may accumulate over time. Anthropic itself puts it unambiguously: “As Claude follows your channel, it continually learns more about the work.” The (relative) good news for teams concerned with compartmentalization is that Claude instances can be separated by department to prevent one team’s data from spreading to another.

In terms of European regulation, companies using Claude Tag under the AI Act have good reason to check their situation: the provisions on general-purpose AI systems (GPAI) have been in force since 2 August 2025, and the use of an AI agent with permanent access to professional communications may fall within the scope of the transparency obligations imposed on deployers.Anthropic, which itself uses an internal version of Claude Tag to generate some 65% of the code changes made by its product teams, is probably the first to have a precise understanding of what this type of tool actually does in a workplace.

Claude Tag is available in beta for Claude Enterprise and Claude Team subscribers. A trial credit period is offered upon activation. Teams that already have the old Claude in Slack integration have until August 3 to migrate at their own pace, after which Anthropic will do so automatically for them.

👉🏻 Follow tech news in real time: add 01net to your Google sources, and subscribe to our WhatsApp channel.

Source: Anthropic