Is Claude Dashboards & Motion Safe? What It Can Reach

Short answer: Claude Dashboards connects to your company’s data and writes queries against it, so the safety question is mostly about access, not about the feature. Anthropic’s launch material is clear on some controls and silent on others. Claude Motion is a different case: it animates things you give it and does not need to reach your databases.

Both launched as betas on 8 October 2026. This post sticks to what is written down in Anthropic’s announcement and its help pages, and says plainly where those pages stop. We have not used either feature, and nothing here is a test result.

What was launched

Claude Dashboards builds a live dashboard from a plain-language question. Anthropic names Amazon Redshift, BigQuery, ClickHouse, Databricks and Snowflake as supported data platforms, plus Salesforce through its connectors. It is in beta on Pro, Max, Team and Enterprise plans, according to the Dashboards help guide.

Claude Motion turns reports, charts and walkthroughs into short animated explainers. Anthropic says Claude writes editable code for the text, shapes, images, charts and timing, and that no video-generation model is involved, so there is no generated footage or AI-generated people. You can download the result as an MP4. Motion is in beta on Team and Enterprise only. See the Motion guide.

The same announcement moved Docs, Slides and Design out of beta onto every plan, including Free. Independent coverage from TestingCatalog matches those details.

What Anthropic documents about control

• —Each chart in a dashboard has its own SQL query. Claude runs it against your data source, and you can open the query to check what a number counts. It also shows when the chart was last refreshed.
• —Dashboards start private to you. Sharing is a separate step: you open the dashboard and click Share.
• —On Enterprise, both features are off by default. Owners switch them on under Organization settings, then Artifacts, and Enterprise owners can limit Dashboards to specific groups with custom roles.
• —Admins control whether people can share beyond the organisation, including by link.
• —Dashboards count toward your plan’s usage limits, like other work with Claude.

Showing the query behind every number is the most useful safety feature here. It lets a person who knows the data catch a wrong join or a filter that quietly drops rows. It does not help if nobody opens the query. If you have never had to check an AI’s work against its source, how to check an AI answer when you are not the expert is a practical start.

What the pages do not say

This is the part that matters for a security review, and the documentation we read does not cover it. Anthropic’s announcement says it does not describe retention policies, detailed access permissions or how connector credentials are stored, and the Dashboards guide does not say whether access to a data source is read-only. We could not confirm any of the following:

• —Whether Claude’s access to a warehouse is read-only, or what permissions the connected account needs.
• —How long dashboard data or query results are kept, and where.
• —Whether results from your data source are used for training. Data handling follows your plan, so start from the terms for that plan rather than assuming.
• —Anything about how Salesforce or warehouse credentials are stored.

On consumer plans, Anthropic’s consumer terms let you choose whether chats are used for training, and say those terms do not apply to Claude for Work or API use. Dashboards need a paid plan, and the data they touch is usually a company’s. That makes it a question for whoever runs your account, not a setting to guess at.

A practical way to try it

1. 01Ask your IT or data team first. Connecting a warehouse is a permissions decision, and it is theirs.
2. 02Use a read-only database account scoped to the tables the dashboard needs. That is a standard precaution on our part, not something Anthropic’s pages promise.
3. 03Start with a small, non-sensitive dataset. Compare three or four numbers against a report you already trust.
4. 04Open the query behind at least one chart before you rely on it.
5. 05Leave sharing off until the numbers have been checked. Dashboards start private for a reason.

For background on what sits underneath a tool like this, AI agent vs LLM explains the difference. If your data lives in spreadsheets rather than a warehouse, how to create a report in Excel and uploading Excel to ChatGPT cover that route and its data question. Anthropic’s other launches this month got the same treatment: is Claude Haiku 5.5 safe and is Claude Sonnet 5.5 safe.

If you used the standalone Claude Design site

One change in the same announcement is not about safety, but it can lose work. Design now lives inside Claude conversations. The standalone site stays available until 14 December 2026. Anthropic’s migration guide says chats and comments there will not carry over, and public links to standalone projects stop working when it closes. Team design systems can be migrated from the Artifacts page.

Coursium is a mobile app that teaches people to use AI at work, and it is on the App Store. Tools like these change every few weeks. Asking where your data goes, checking the output against the source and keeping sharing narrow carry over to whatever ships next. If that is what you want to practise, have a look at Coursium.