2018787 - Track security flags for automatically added initial chat content

Status: RESOLVED FIXED

(Core :: Machine Learning: General, task)

Description

[Simon Friedberger [:simonf]]

We currently proactively provide the open tab and - if there are matching ones - memories about the user to the LLM. This requires setting the tracking flags correctly.

Comment 1

[Randy Concepcion]

Attachment: Bug 2018787 - Set security flags when initial chat content is injected r=simonf

Sets conversation securityProperties (untrusted_input, private_data) when
initial chat content is automatically injected in generatePrompt(). Tab
metadata from content pages (http/https) is treated as untrusted and
private. Internal pages (about:, chrome:) stay clean unless memories or
@mentions are also injected. Flags are monotonic and only set when content
is actually included. Tracking only, no tool gating changes.

Comment 2

[Simon Friedberger [:simonf]]

Attachment: Bug 2018787: Init security flags on conversation start r=thasan

Comment 3

[Simon Friedberger [:simonf]]

Attachment: Bug 2018787: Make getRealTimeInfo static r=thasan

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/a70cff25acb0
https://hg.mozilla.org/integration/autoland/rev/00ae018ef9ec
Init security flags on conversation start r=thasan,ai-frontend-reviewers,rconcepcion,jlewis
https://github.com/mozilla-firefox/firefox/commit/b70d3d1e067c
https://hg.mozilla.org/integration/autoland/rev/510c1169ded4
Make getRealTimeInfo static r=thasan,ai-frontend-reviewers,gcardieri

Comment 5

[Silaghi Andreea]

Backout for causing xpcshell failures at ChatConversation.js

Backout link
Push with failure
Failure log
Failure log for TV
Failure line TEST-UNEXPECTED-FAIL | browser/components/aiwindow/ui/test/xpcshell/test_ChatConversation.js | xpcshell return code: 0

Comment 6

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/bbe0c99ebbc1
https://hg.mozilla.org/integration/autoland/rev/6e7412772dd1
Init security flags on conversation start r=thasan,ai-frontend-reviewers,rconcepcion,jlewis

Comment 7

[Cristian Tuns]

https://hg.mozilla.org/mozilla-central/rev/6e7412772dd1

Comment 8

[Greg Tatum [:gregtatum]]

[Tracking Requested - why for this release]:
We'll want this for 150 to make sure security stuff for smart window chats works correctly

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/8ba771c8cfb8
https://hg.mozilla.org/integration/autoland/rev/b47418c1b714
Make getRealTimeInfo static r=thasan,ai-frontend-reviewers,gcardieri

Comment 10

[rperta]

https://hg.mozilla.org/mozilla-central/rev/b47418c1b714

Comment 11

[Greg Tatum [:gregtatum]]

Attachment: Bug 2018787: Init security flags on conversation start

Original Revision: https://phabricator.services.mozilla.com/D288663

Comment 12

[Phabricator Automation]

firefox-beta Uplift Approval Request

• User impact if declined/Reason for urgency: This is a required security behavior for the Smart Window MVP.
• Code covered by automated testing?: yes
• Fix verified in Nightly?: yes
• Needs manual QE testing?: no
• Steps to reproduce for manual QE testing:
• Risk associated with taking this patch: low
• Explanation of risk level: This risk is lower, as the feature still needs to be turned on.
• String changes made/needed?: None
• Is Android affected?: no

Comment 13

[Greg Tatum [:gregtatum]]

Attachment: Bug 2018787: Make getRealTimeInfo static

Original Revision: https://phabricator.services.mozilla.com/D288860

Comment 14

[Pulsebot]

https://github.com/mozilla-firefox/firefox/commit/e6d26d56ebb2
https://hg.mozilla.org/releases/mozilla-beta/rev/cf40e1061045
https://github.com/mozilla-firefox/firefox/commit/22426fdcac05
https://hg.mozilla.org/releases/mozilla-beta/rev/fbbb3fa55209