If you run Claude Enterprise, Claude Data switches on for your whole organization on November 19, 2026, unless an owner saves a different setting first. The default itself is mild: every viewer still queries through their own connections. The setting that needs a decision is live sharing, a beta that lets a colleague see warehouse results through your credentials even when they have no access to the source.
Anthropic launched Claude Data (the renamed Claude Dashboards), Claude Sheets, Claude Whiteboards and interactive connectors at a San Francisco event on November 10, as Fortune reported, on the same day its IPO roadshow began. Microsoft shares fell more than 2% on the news, per the same report. For an admin, the launch comes down to one settings page and nine days to fill it in.
What Changes on November 19?
On November 19, Claude Data moves from off to on for every Enterprise organization whose owner has not already saved a choice. Anthropic's Claude Data support article says it plainly: "On Enterprise plans, Claude Data turns on by default on November 19, 2026," and owners can restrict it to specific groups with custom roles. The two beta features inside it, certified queries and live sharing, stay off by default on Enterprise.
Team plans get a different deal. On Team, Claude Data and both betas are already on. If you run a Team workspace for a business unit alongside an Enterprise tenant, the two now behave differently from today.
Claude Data is a Claude feature that runs queries against your connected data sources and turns the results into dashboards, scrolling data stories, narrated short videos or written analyses. The support article names Amazon Redshift, BigQuery, ClickHouse, Databricks and Snowflake as warehouses, plus connected apps such as Salesforce. It is available on Pro, Max, Ultra, Team and Enterprise, and it draws on your plan's usage limits like any other Claude work.
Two operating details matter for whoever owns BI. Claude Data "doesn't refresh on a schedule"; it re-queries when someone opens a dashboard whose data is due. And each chart shows its query and when it last refreshed. A dashboard nobody opens for a month shows month-old numbers, with the timestamp right there, and that is fine for a quick look but wrong for anything a finance team reports from. Anthropic's own guidance in the same article is that dashboards suit quick exploration and deeper analysis belongs in another tool.
The controls live in Organization settings > Artifacts, under Templates, with certified queries and live sharing under Data. Anthropic's Artifacts admin guide confirms the November 19 date and adds that both beta toggles are hidden while the Data template is off.
Live Sharing Lets Viewers See Data They Cannot Query
Live sharing is the one control here that can change who sees what in your warehouse, so it should stay off until the data owner signs off. Without it, Anthropic says, each viewer's dashboard loads data through that viewer's own connections. Row-level security in Snowflake or Databricks keeps working because the viewer's identity is the one hitting the warehouse.
With live sharing on, the model flips. The support article says "Live sharing runs on your access to the data source," so people in your organization see the results even without access to the source. Anthropic limits the exposure to the queries the dashboard needs, and those queries cannot change without an editor's approval. That is a reasonable design for a sales VP sharing pipeline numbers with a team that has no Salesforce seats. It is also a way for a well-meaning analyst with broad warehouse rights to publish compensation or customer data to people your access reviews would never approve.
Your warehouse governance probably assumes the identity that runs a query is the identity that sees the answer. Live sharing breaks that assumption for anyone the sharer chooses. We covered the same pattern in agentic analytics work at Grab, where governed metric definitions mattered more than the model, and in our comparison of Snowflake, Databricks and ThoughtSpot text-to-SQL tools, where row-level security decided which product fit.
If you do enable live sharing, scope it with custom roles to the groups whose warehouse grants are already narrow, and tell the data owners which service accounts or user identities will be doing the sharing.
Certified Queries Need an Owner Before They Need a Badge
Certified queries are only as trustworthy as the people allowed to certify them, and by default that is anyone with write access to the artifact. A certified query is a query your data team approves so Claude can reuse it, and certified charts show a badge. Per the support article, "Anyone with write access to a Data artifact can certify," and editing a certified query removes its certification.
A badge any editor can grant does not tell a reader much. Anthropic offers a stricter path: organizations that keep certified queries in a Git-backed plugin manage them through GitHub pull requests, and certified queries can also draw on dbt, Looker or Cube definitions through existing MCP connectors. If you already have a semantic layer, point Claude Data at it and route certification through the same review your dbt models get. Our piece on the UNICEF MCP benchmark showed how badly a generic connector does without curated definitions, which is the gap certified queries are meant to close.
Sheets, Whiteboards and Connectors Stay Off Until You Act
Claude Sheets and Claude Whiteboards are betas that Enterprise owners must switch on, so they do not need action before November 19. The Artifacts admin guide lists both as on by default for Team and off on Enterprise "until an owner turns them on." Sheets files export as .xlsx or CSV, per iClarified's launch summary, which also notes that Sheets has only basic spreadsheet functions for now.
The same admin guide has two sharing settings worth checking while you are on that page. Email invitations to people outside your organization are off by default on Enterprise and on for Team, and pending invitations expire after 30 days. External link sharing lets users share artifacts with anyone who has the link; the guide does not state its default, but it does say turning it off also stops existing public links unless an artifact is allowed individually. Check the current value in your tenant rather than assuming it.
Interactive connectors are the other half of the launch. Anthropic's launch post says partners such as Adobe and Atlassian can now embed their apps directly in Claude, that plugins from Salesforce (beta), NetSuite and Workday (beta) come with built-in skills, and that the Claude Marketplace lists more than 9,000 connectors and plugins. It says "Claude works within your existing permissions," and that admins can limit actions so that, in Anthropic's example, Claude can read Slack but not send messages. Claude auto-connect is in beta on eligible Team and Enterprise plans, and connects employees the first time they use a tool. Decide before you enable auto-connect whether you want connectors to spread by first use or by request.
Is Claude Meetings Live? The Reports Disagree
Claude Meetings has no Anthropic documentation yet, so treat its availability as unconfirmed. Fortune reported that Meetings is in beta for Team and Enterprise plans, starting with Google Meet, and joins calls as a "visible participant." Anthropic's own launch post says "Coming soon, Claude Meetings will let you add Claude to a call to take notes," and The Decoder also describes it as not yet launched.
Either way, write the meeting-bot policy now, before the toggle exists. Google Meet has its own lever: Meet safety settings can require participants to consent before recording, transcription or Gemini note-taking, and can remove people who decline. Google's page describes that control in terms of Meet's own features and does not say whether it covers a third-party participant like Claude, so test it with a pilot group once Meetings reaches your tenant. If your legal team requires all-party consent for recordings in some jurisdictions, that rule applies to an AI participant too.
The Steel-Man: Why Turning It On Early Makes Sense
The strongest case for enabling Claude Data before November 19 is control: you pick the groups, rather than inheriting an org-wide default. Anthropic's Artifacts admin guide says owners can enable it sooner, and the support article says the default only applies if no owner has saved a setting. Turning it on today for a pilot group of analysts and finance partners, with live sharing off, gives you nine days of real usage data before the rest of the company arrives.
The default-off settings are also the safe ones. With live sharing and certified queries off, a Claude Data dashboard is a chart built from queries each viewer could already run, so the exposure sits in the two betas. This is the same pattern OpenAI used with its ChatGPT Enterprise Dots default-on date: an admin who acts early decides the scope, and one who waits gets the vendor's.
The case against is overlap. If your company already pays for a BI platform with governed metrics, Claude Data creates a second place where numbers get published, refreshed only when opened, with certification any editor can grant. Scoping it to the teams without BI seats handles most of that.
What to Do Before November 19
The deadline is the default flip on November 19; everything below can start this week.
This Week:
- Open Organization settings > Artifacts in your Claude Enterprise admin console and save an explicit Claude Data setting, on for named groups or off, so the November 19 default does not decide for you.
- Confirm live sharing and certified queries are off, and write down who must approve turning either on. The data platform owner should be one of them.
- List which warehouse and CRM connectors are already enabled in your tenant. Those are the sources Claude Data can query on day one.
Before November 19:
- Run a pilot group on Claude Data with live sharing off. Ask the warehouse team to check query logs for unexpected load, since dashboards re-query on open.
- Check your external link sharing setting under Artifacts and record its current value.
- If you use dbt, Looker or Cube, decide whether certification will go through the Git-backed plugin and pull requests instead of the per-artifact button.
This Month:
- Draft a meeting-bot policy covering Claude Meetings, consent and retention, and test Google Meet's consent settings against it once Meetings appears in your admin console.
- Review connector auto-connect and action limits before enabling Sheets or interactive connectors broadly. Our earlier pieces on Claude Enterprise inference hooks and DLP and Smart Reports privacy settings cover the adjacent controls.
- Ask your Anthropic account team whether audit logs and the Compliance API record live-sharing events and certification changes. The Compliance API documentation does not mention artifact sharing.
The Bottom Line
Vendors now ship workplace AI features to Enterprise tenants on dated default-on schedules, and the admin who reads release notes sets the scope. Anthropic did the same with Docs, Slides and Design, which the Artifacts admin guide shows turned on by default for Enterprise on October 15. Claude Data follows that pattern with a safer base setting and one beta, live sharing, that changes the identity behind a query. Claude also now reaches into Google Docs, Sheets and Slides through an add-on that, per Anthropic's October 6 post, follows existing Google sharing permissions. In each of these features the base setting uses the user's own access, and the toggles that need review are the ones that let one user's access stand in for another's.
Save the Claude Data setting before November 19, and keep live sharing off until the data owner has approved it in writing.
Continue Reading
- Your Claude Enterprise DLP Hooks May Let Unchecked Prompts Through
- Set Privacy Rules Before Enabling Claude Enterprise Smart Reports
- OpenAI Starts Billing for ChatGPT Enterprise Dots on November 18
- Personal Claude Accounts Expose Work Laptops to ThisBash
- Grab Ran Analytics Agents 5 Months. Context Was the Ceiling.
- Snowflake Copilot vs Genie vs ThoughtSpot: Copilot Is the Wrong Tool
