Sally Security Concerns - PSA

A quick update:

• We are planning to provide an admin setting to know which channels should not be supported. We can then make this check every time and block both conversations (querying or engaging with Sally) and notifications (alerts pushed to such channels).
• For channels that are not blocked, we are going to change the base interaction to include a warning message + consent before posting anything. We looked into showing responses privately as well (called ephemeral messages), but are running into some text size limitations.

We are making these changes as quickly as possible, and will keep you all posted on the progress here. Furthermore, we are doing a similar assessment for Sally in MS Teams as well.

This might be less feasible from your side, but given that we have 26k Slack channels and new ones would be impossible to monitor, could it be an opt in process rather than an opt out process?

Manu, might I suggest the reverse? We cannot consistently maintain a block list, but an allow list is far more reasonable. 

If we do that, I'm assuming we would want to block only external channels by default? Otherwise, it'll be significant overhead to keep adding all the internal channels as well. If yes, that's where things can get tricky. When Sally is invoked from a channel, we don't get information on whether or not that channel is external. So we'd have to do something like this: 

1. Preload and save external channels on our end as a block list.
2. Any external channel added to an admin's allow list gets removed from #1.

I'll discuss this with Engineering in more detail.

I think by default any external channel would need blocking.

In general, if the solution is to mitigate possible data issues, if I have to opt out every channel I don’t want it on I’d need an admin just to play whack a mole with Slack channels to find out if it needs to be blocked or not and is a solution that doesn’t really address the problem.

If long term we can bulk manage channels that’s great, but for short term use we’d want to start with an opt in if it’s a channel by channel thing.

Makes sense, Bradley. We don’t want to add unnecessary overhead. Let me get some more details and also understand effort implications. If we have to preload external channels and maintain a block list, we have what we need at that point to automatically block all external channels. Step #2 adds the ability to selectively use Sally in external channels as and when it makes sense.

Team,

Here’s where we landed:

• We are still good with this part: “For channels that are not blocked, we are going to change the base interaction to include a warning message + consent before posting anything. We looked into showing full responses privately as well (called ephemeral messages), but are running into some text size limitations.”  We are adding an admin setting as well to decide whether this happens in all channels or just public ones (default behavior).
• We will also allow blocking Sally entirely based on three options (multi-select, so can apply all three for tight control):
  1. Block in public channels
  2. Block in external channels (with connected external organizations)
  3. Block in channels with the following prefixes… (e.g., “cust_” based on naming best practices)

Let me know please if there are any questions. Appreciate your patience and partnership!

Thanks,

Manu

Thanks for this Manu - do you have an ETA on when these items will be deployed?

3-4 weeks roughly. Will keep this group posted.