Skip to main content

Product Office Hours #13 - Sensitive Data Restrictions - Data Security for MacOS - Nov 12th *Q&A in thread*

  • Nov 6, 2026 2:00 PM UTC
  • 3 replies
  • 81 views

Mitchell_Gordon
Forum|alt.badge.img+11

Hi Nation,

Product Office Hours #13 - Sensitive Data Restrictions - Data Security for MacOS

Next session: Thursday, 12th Nov -  7 am PST / 9 am CST / 3 pm GMT / 4 pm CET 
Speakers: Hamilton Yang, Product Manager
Register here: https://jamf.it/ProductOfficeHours

 
🧵 Got a question? Drop it in the comments below ahead of time if you want it on our radar, though you don't have to. We'll be taking questions live on the call! The Q&A transcript will be added to this thread post event.

See you there!

3 replies

mvu
Forum|alt.badge.img+25
  • Jamf Heroes
  • Nov 6, 2026 4:44 PM UTC

👍


mattjerome
Forum|alt.badge.img+16
  • Jamf Heroes
  • Nov 6, 2026 7:52 PM UTC

Looking forward to this one.


LysetteB
Forum|alt.badge.img+20
  • Community Manager
  • Nov 17, 2026 4:57 PM UTC

This week's Product Office Hours covered Sensitive Data Restrictions and data security for macOS, with Hamilton Yang (Principal Product Manager). Here's every question asked during the session, pulled from both the live Q&A and the chat. 

Is the beta running now, and what will participants be involved in? 

The beta is already underway, but more participants are needed. The ideal is for testers to put the product through its paces and give plenty of feedback, and they can meet as often as they like. At a minimum, expect a monthly check-in to talk through how the beta is going and keep the two-way conversation going. GA is roughly four to six months away. It could come sooner if early testers like what they see, but getting the policy-building experience right will take a few months. 

To take part in the beta, email hamilton.yang@jamf.com 

Is there a roadmap to use Microsoft Purview sensitivity labels, so we could remove Defender? 

This is actively in progress. Microsoft Office files carry Purview labels inside their metadata, and the team is working on extracting those labels so you can write policies around them. That capability is expected to arrive in the beta shortly, roughly a week after the session. Third-party file formats, such as CAD files, will likely remain a gap, because Purview tags on those files don't travel with the file itself. Closing that gap would probably need an integration with a Purview server, and that isn't designed yet, so we won't over-promise on it. 

Why use Claude in the demo rather than mdfind? 

mdfind would trigger the same alert, so it's a valid route too. Claude was used in the demo to show a more advanced case. AI tools often spawn several layers of child processes, and the solution traces the full process ancestry back to the originating app. The team is also checking whether mdfind is excluded by default as a system process. If you have a view on whether mdfind events would be too noisy to report, share it in the beta feedback. 

Is Jamf for Mac required, and do you need Jamf Protect for this? 

The plan is to make this part of Jamf for Mac at no additional cost, although the marketing plan isn't finalised, so treat that as direction rather than a confirmed commitment. If you don't have Jamf Protect set up today, you'll need access to the Jamf Protect console to use the feature, but you don't need to deploy the Jamf Protect agent. There's a separate SDR-specific agent or package, so until you're ready to use Sensitive Data Restrictions, you don’t have to deploy that. 

Can it block or kill processes? For example, can it stop an AI tool from reading sensitive files? 

It can block that process from accessing files, but it doesn't kill the process. Blocking sensitive data acccess is available now, so a rule can stop a process, such as anything signed by OpenAI and its child processes, from touching certain files. The user sees the app report that it doesn't have permission to access the file, and that's all they see – the app continues to run. This works best alongside application-layer controls such as Jamf AI governance. AI governance covers the harness and prompt level, while Sensitive Data Restrictions adds the OS-level layer, including USB and AirDrop controls that AI governance doesn't cover. Better integration between the two is on the roadmap. 

How does it handle encrypted files and folders, and USB drives that are encrypted with third-party tools? 

Apple's built-in USB encryption can be detected, so you can set a policy that only allows files to move to USB volumes with that encryption enabled. Third-party encryption that is transparent, where you still access the volume through Finder after decryption, should work. FileVault and client-side encryption also work, as long as the encrypted volume is mounted. Encryption that needs a separate third-party app or custom utility to decrypt the contents is unlikely to be inspectable, and that is probably by design for those encryption vendors, and not supported by us. 

Can USB storage rules allow only approved USB keys, identified by GUID, and only when they are encrypted with the recovery key held in Jamf? 

We support scoping rules down to the serial ID (as well as vendor ID and product ID). If a volume decrypts, we can control it – but we don’t integrate into keychains/credentials and we don’t differentiate based on where that credential was sourced from. If you’ve got a use case here, help us design it! 

Is anyone in the beta already exploring DLP policies for MNPI related to other organisations? 

Not yet. Jamf is looking for help from the community to define sensible default rules, packages and content. The rules engine is currently bare-bones, and there's a lot of polish still to come. The best way to shape it is to join the beta and say what rules, packages and content you need. 

Can it detect screenshots that contain sensitive data? 

We believe so. Sensitive Data Restrictions uses the Spotlight index, and Spotlight already performs OCR on screenshots, so sensitive text in a screenshot should be picked up. Jamf's own testing looks solid, but this is an assumption that still needs validating in the beta. Building on what Apple already provides also avoids adding extra computational overhead on the device. 

How does it handle medical images? 

This wasn't covered in the session. Please share more detail about the file types and workflows involved so this can be looked into. But directionally, we think local AI models could help. 

What exactly is the beta? Is it a DLP product, a security scanner or a data governance tool? 

It's endpoint DLP for macOS. There are no plans for iOS yet, as the priority is getting it right on the Mac first.