Page MenuHomePhabricator
Create Task
Maniphest T430772

DE4.1.3: Integrate personal information detection classifier into production
Open, Needs TriagePublic
Actions

Description

Summary

In WE4.12 Content policy model evaluation, we worked with community members to identify likely suppressable or revision deletable edits based on community defined policy in https://en.wikipedia.org/wiki/Wikipedia:Oversight. In this epic, we will bring this proof of concept into on-wiki workflows

Details

Other Assignee
kostajh

Related Objects
Search...

StatusSubtypeAssignedTask
 OpenDreamy_JazzT430772 DE4.1.3: Integrate personal information detection classifier into production
 ResolvedMPostoronca-WMFT431289 Update the special page displaying content policy to use the first version of the design
 ResolvedROdonoghue-WMFT431288 Design: Create mocks for a special page to display flagged items
 ResolvedDreamy_JazzT433367 Create initial special page for displaying content policy violations
 ResolvedMPostoronca-WMFT432931 Allow authorized users to add or remove the false positive tag
 ResolvedMPostoronca-WMFT431292 Tag revisions with private change tag for likely oversightable/revision deletable content
 ResolvedFeatureDreamy_JazzT386456 Feature: Allow for private, rights-restricted tags
 ResolvedDreamy_JazzT431414 Create ChangeTagsFormatter service
 OpenNoneT431523 Update Sqoop to get `change_tag` and `change_tag_def` from production replicas
 OpenGGoncalves-WMFT431524 Remove private tags from mediawiki-history public dumps
 ResolvedMPostoronca-WMFT431667 Update ChangeTagsStore database query methods for rights-restricted tags
 ResolvedDreamy_JazzT431696 Drop hard-deprecated ChangeTag methods deprecated in 1.41
 ResolvedDreamy_JazzT431873 Update 'revisions' query API to not allow filtering for revisions based on private tags the user cannot see
 ResolvedDreamy_JazzT432172 Special:PendingChanges does not prevent users from filtering on private tags they cannot see
 ResolvedDreamy_JazzT431287 Define a "false positive (personal information)" private change tag
 ResolvedDreamy_JazzT431295 Define a change tag for "likely contains personal information"
 ResolvedMPostoronca-WMFT432219 Content policy: Create a hook to handle results from the CoPE model and decide consequences to take
 ResolvedMPostoronca-WMFT431286 Call CoPE models for edits that are made
 ResolvedkostajhT431023 Create and deploy Extension:WikimediaAntiAbuse
 ResolvedkostajhT432283 Application Security Review Request : WikimediaAntiAbuse
Restricted Task
Restricted Task
 ResolvedMPostoronca-WMFT431291 Implement mechanism for deciding which edits are checked for a given policy
 ResolvedkostajhT432457 Create service to generate snippets for the context header and content for calling a CoPE model
 DuplicateNoneT431293 Define mapping of private change tag for likely suppressable content with user rights
 ResolvedDreamy_JazzT432230 Create a maintenance script to evaluate a content policy text against arbitary content
 ResolvedDreamy_JazzT432229 Create a service to call a CoPE model with given policy text for given content
Restricted Task
 ResolvedMPostoronca-WMFT432452 Echo notifications when new items are flagged
 ResolvedMPostoronca-WMFT432848 EventLogging: Record confidence score results
 ResolvedDreamy_JazzT434716 Special:AbuseReview does not show revisions from deleted pages
 ResolvedkostajhT435020 AbuseReview: Add ability to indicate "no further action needed" for revisions
 OpenNoneT435516 AbuseReview: Create Superset dashboard
 OpenDreamy_JazzT435517 AbuseReview: Server side instrumentation
 OpenDreamy_JazzT435519 AbuseReview: Implement design for filtering results
 OpenDreamy_JazzT436017 AbuseReview: Add username filter to filter dialog
 ResolvedDreamy_JazzT436014 AbuseReview: Use a dialog for filtering the results
 OpenMPostoronca-WMFT435901 Hide personal info notifications when the revision has been marked "no further action"
 OpenMPostoronca-WMFT435925 AbuseReview: Implement attribution for verdicts
 OpenkostajhT435983 Update Special:AbuseReview with table header and inline verdicts design
 OpenMPostoronca-WMFT436063 AbuseReview: Implement info button to show more information about verdicts and flags
 OpenkostajhT436490 AbuseReview: Use core's diff libraries with inline format
 OpenNoneT436491 AbuseReview: Update row status after suppression action
Restricted Task
 OpenMPostoronca-WMFT436492 AbuseReview: Client-side instrumentation
 OpenDreamy_JazzT436546 AbuseReview: Support deep linking to revision or set of revisions
 OpenDreamy_JazzT436958 AbuseReview: Instrument views of AbuseReview through echo notifications
 OpenNoneT436656 AbuseReview: Change "Needs further action" to "Send back for review"
 OpenNoneT436912 Review personal info classifier false positives with community members and refine prompt

Event Timeline

kostajh created this task.Jul 1 2026, 7:26 AM
Restricted Application added a subscriber: Aklapper. · View Herald TranscriptJul 1 2026, 7:26 AM
kostajh assigned this task to Dreamy_Jazz.Jul 1 2026, 7:27 AM
kostajh updated the task description. (Show Details)
kostajh moved this task from Backlog to Epics in progress on the Product Safety and Integrity (Sprint Venus Flytrap (Jun 29 - Jul 17)) board.Jul 1 2026, 7:36 AM
kostajh updated Other Assignee, added: kostajh.Jul 1 2026, 7:51 AM
Barkeep49 subscribed.Jul 2 2026, 2:58 PM
Xaosflux subscribed.Jul 2 2026, 8:54 PM
asilvering subscribed.Jul 3 2026, 1:42 AM
kostajh added a subtask: T431023: Create and deploy Extension:WikimediaAntiAbuse.Jul 6 2026, 12:26 PM
kostajh added a subtask: T386456: Feature: Allow for private, rights-restricted tags.
kostajh updated the task description. (Show Details)
kostajh created subtask T431286: Call CoPE models for edits that are made.Jul 6 2026, 12:32 PM
kostajh created subtask T431287: Define a "false positive (personal information)" private change tag.
kostajh created subtask T431288: Design: Create mocks for a special page to display flagged items.Jul 6 2026, 12:36 PM
kostajh created subtask T431289: Update the special page displaying content policy to use the first version of the design.
kostajh changed the status of subtask T431289: Update the special page displaying content policy to use the first version of the design from Open to Stalled.
kostajh created subtask T431291: Implement mechanism for deciding which edits are checked for a given policy.Jul 6 2026, 12:41 PM
kostajh created subtask T431292: Tag revisions with private change tag for likely oversightable/revision deletable content.Jul 6 2026, 12:44 PM
kostajh created subtask T431293: Define mapping of private change tag for likely suppressable content with user rights.Jul 6 2026, 12:50 PM
kostajh created subtask T431295: Define a change tag for "likely contains personal information".Jul 6 2026, 12:52 PM
Johannnes89 subscribed.Jul 7 2026, 5:19 PM
Barras subscribed.Jul 7 2026, 5:19 PM
DerHexer subscribed.Jul 7 2026, 5:24 PM
Amorymeltzer subscribed.Jul 8 2026, 4:59 PM
kostajh changed the status of subtask T431292: Tag revisions with private change tag for likely oversightable/revision deletable content from Open to Stalled.Jul 10 2026, 11:16 AM
kostajh mentioned this in T432283: Application Security Review Request : WikimediaAntiAbuse.Jul 15 2026, 6:32 PM
sbassett awarded a token.Jul 15 2026, 6:37 PM
Dreamy_Jazz added a subtask: T432230: Create a maintenance script to evaluate a content policy text against arbitary content.Jul 15 2026, 11:43 PM
kostajh added a subtask: Restricted Task.Jul 16 2026, 3:12 PM
Dreamy_Jazz changed the status of subtask T431292: Tag revisions with private change tag for likely oversightable/revision deletable content from Stalled to Open.Jul 16 2026, 5:02 PM
Dreamy_Jazz closed subtask Restricted Task as Resolved.Jul 16 2026, 10:42 PM
kostajh created subtask T432452: Echo notifications when new items are flagged.Jul 17 2026, 11:40 AM
kostajh renamed this task from DE4.1.1: Integrate oversight content policy model into production to DE4.1.1: Integrate personal information detection model into production.Jul 17 2026, 3:56 PM
OKryva-WMF edited projects, added: Product Safety and Integrity (Sprint Clover (Jul 20 - Aug 7)); removed: Product Safety and Integrity (Sprint Venus Flytrap (Jun 29 - Jul 17)).Jul 20 2026, 1:56 PM
OKryva-WMF moved this task from Backlog to Epics in progress on the Product Safety and Integrity (Sprint Clover (Jul 20 - Aug 7)) board.
kostajh closed subtask T431023: Create and deploy Extension:WikimediaAntiAbuse as Resolved.Jul 22 2026, 11:30 AM
kostajh renamed this task from DE4.1.1: Integrate personal information detection model into production to DE4.1.1: Integrate personal information detection classifier into production.Jul 22 2026, 11:42 AM
kostajh created subtask T432848: EventLogging: Record confidence score results.
kostajh created subtask T432931: Allow authorized users to add or remove the false positive tag.Jul 23 2026, 7:52 AM
Dreamy_Jazz closed subtask T432230: Create a maintenance script to evaluate a content policy text against arbitary content as Resolved.Jul 27 2026, 3:25 PM
Dreamy_Jazz closed subtask T431295: Define a change tag for "likely contains personal information" as Resolved.Jul 27 2026, 3:38 PM
Dreamy_Jazz closed subtask T431287: Define a "false positive (personal information)" private change tag as Resolved.
Dreamy_Jazz closed subtask T431286: Call CoPE models for edits that are made as Resolved.Jul 27 2026, 3:41 PM
Dreamy_Jazz closed subtask T431291: Implement mechanism for deciding which edits are checked for a given policy as Resolved.
Dreamy_Jazz changed the status of subtask T431289: Update the special page displaying content policy to use the first version of the design from Stalled to Open.Tue, Aug 11, 1:24 PM
OKryva-WMF edited projects, added: Product Safety and Integrity (Sprint Sunflower (Aug 11 - Aug 28)); removed: Product Safety and Integrity (Sprint Clover (Jul 20 - Aug 7)).Tue, Aug 11, 1:52 PM
OKryva-WMF moved this task from Backlog to Epics in progress on the Product Safety and Integrity (Sprint Sunflower (Aug 11 - Aug 28)) board.
kostajh added a subtask: T434716: Special:AbuseReview does not show revisions from deleted pages.Sun, Aug 16, 6:27 AM
kostajh created subtask T435020: AbuseReview: Add ability to indicate "no further action needed" for revisions.Sun, Aug 16, 6:31 AM
kostajh created subtask T435516: AbuseReview: Create Superset dashboard.Thu, Aug 20, 3:43 PM
kostajh created subtask T435517: AbuseReview: Server side instrumentation.Thu, Aug 20, 3:48 PM
kostajh created subtask T435519: AbuseReview: Implement design for filtering results.Thu, Aug 20, 3:55 PM
kostajh closed subtask T432452: Echo notifications when new items are flagged as Resolved.Mon, Aug 24, 12:52 PM
kostajh created subtask T435901: Hide personal info notifications when the revision has been marked "no further action".Tue, Aug 25, 7:25 AM
kostajh created subtask T435925: AbuseReview: Implement attribution for verdicts.Tue, Aug 25, 10:53 AM
kostajh renamed this task from DE4.1.1: Integrate personal information detection classifier into production to DE4.1.3: Integrate personal information detection classifier into production.Tue, Aug 25, 11:41 AM
kostajh created subtask T435983: Update Special:AbuseReview with table header and inline verdicts design.Tue, Aug 25, 4:38 PM
Dreamy_Jazz closed subtask T434716: Special:AbuseReview does not show revisions from deleted pages as Resolved.Wed, Aug 26, 8:37 AM
kostajh created subtask T436063: AbuseReview: Implement info button to show more information about verdicts and flags.Wed, Aug 26, 8:59 AM
kostajh created subtask T436065: AbuseReview: Implement inline revert action.Wed, Aug 26, 9:02 AM
kostajh closed subtask T435020: AbuseReview: Add ability to indicate "no further action needed" for revisions as Resolved.Wed, Aug 26, 10:59 AM
kostajh closed subtask T432931: Allow authorized users to add or remove the false positive tag as Resolved.Wed, Aug 26, 1:18 PM
kostajh closed subtask T431289: Update the special page displaying content policy to use the first version of the design as Resolved.Wed, Aug 26, 1:44 PM
kostajh closed subtask T431292: Tag revisions with private change tag for likely oversightable/revision deletable content as Resolved.Thu, Aug 27, 7:45 AM
kostajh closed subtask T386456: Feature: Allow for private, rights-restricted tags as Resolved.
kostajh closed subtask T431288: Design: Create mocks for a special page to display flagged items as Resolved.Thu, Aug 27, 9:09 AM
kostajh created subtask T436490: AbuseReview: Use core's diff libraries with inline format.Mon, Aug 31, 6:07 AM
kostajh created subtask T436491: AbuseReview: Update row status after suppression action.Mon, Aug 31, 6:15 AM
kostajh added a subtask: Restricted Task.Mon, Aug 31, 6:20 AM
kostajh created subtask T436492: AbuseReview: Client-side instrumentation.Mon, Aug 31, 6:27 AM
kostajh created subtask T436546: AbuseReview: Support deep linking to revision or set of revisions.Mon, Aug 31, 6:03 PM
Dreamy_Jazz closed subtask T432848: EventLogging: Record confidence score results as Resolved.Tue, Sep 1, 1:03 PM
kostajh created subtask T436656: AbuseReview: Change "Needs further action" to "Send back for review".Tue, Sep 1, 2:06 PM
OKryva-WMF edited projects, added: Product Safety and Integrity (Sprint Heather (Aug 31 - Sept 18)); removed: Product Safety and Integrity (Sprint Sunflower (Aug 11 - Aug 28)).Wed, Sep 2, 11:21 AM
OKryva-WMF moved this task from Backlog to Epics in progress on the Product Safety and Integrity (Sprint Heather (Aug 31 - Sept 18)) board.
kostajh created subtask T436912: Review personal info classifier false positives with community members and refine prompt.Thu, Sep 3, 9:43 AM
kostajh removed a subtask: T436065: AbuseReview: Implement inline revert action.Thu, Sep 3, 10:55 AM
Content licensed under Creative Commons Attribution-ShareAlike (CC BY-SA) 4.0 unless otherwise noted; code licensed under GNU General Public License (GPL) 2.0 or later and other open source licenses. By using this site, you agree to the Terms of Use, Privacy Policy, and Code of Conduct. · Wikimedia Foundation · Privacy Policy · Code of Conduct · Terms of Use · Disclaimer · CC-BY-SA · GPL · Credits