Translator from classification scores to customer moderation actions

Checkstep

Checkstep provides moderation workflows that combine scanner outputs, customer policies and human review. In its documented integration with Sightengine, classification scores become inputs to decisions that a customer platform can implement through event notifications.

Workflow and data permissions

Published integration documentation describes decisions to remove, restore or keep content down, and human author decisions to suspend, terminate or restore accounts. Customers must implement the event handler. Terms identify the customer as controller and Checkstep as processor; clause 5.5 permits customer-data use for service machine learning subject to an email opt-out. These published capabilities and permissions do not establish a particular customer's settings, accepted contract or training use. The documented integration is not a shared cross-customer offender register.

What the records show

Customer platform / subscriber roleCheckstep

retains controller role under published terms

Effective interval unknown; public record observed 9 September 2026.

Terms 14 identify contracting party as controller,Checkstep processor; instructions/protected subprocessors/referred requests. Termination 30 days is backup-request window, not universal deletion deadline. Executed terms unknown.

CheckstepCustomer platform / subscriber role

can return content and account action instructions

Effective interval unknown; public record observed 9 September 2026.

decision can remove/restore/keep content down; human author-decision can suspend/terminate/restore accounts, with content removal depending on decision. Customer must implement handler. Capability, not observed enforcement.

CheckstepCheckstep customer-data training permission

publishes opt-out customer-data ML permission

Effective interval unknown; public record observed 9 September 2026.

Terms 5.5 permit Customer Data for service ML and treat it as customer instruction, subject to email opt-out. No anonymization condition in clause; actual training, pooling, acceptance/opt-out unknown.

Sightengine (Kozelo SAS)Checkstep

describes supplying classification outputs

Currentguide observed 9 September 2026; historical version unknown. PUBLICpublished 26 March 2026 citesguideas 2025.

Coauthored guide describes scores entering Checkstep workflow; PUBLIC calls example signal sharing. Not shared offender list, universal no-external-processing guarantee or observed customer deployment.

CheckstepCustomer platform / subscriber role

can return appeal and restoration events

Effective interval unknown; public record observed 9 September 2026.

incident-closed includes overturned/upheld; restoration and notification events described. No successful appeal, independent adjudication or correction across unrelated platforms established.

From the investigation

Result

PUBLIC's example is documented as an integration of moderation vendors. The observed path is a customer submitting content for analysis, Sightengine returning classification scores, Checkstep applying the customer's configured policy, and a decision event returning to that customer's platform. The records do not establish a shared list of restricted people or automatic propagation of one platform's ban to another.

Read the research & sources ↗
Result

The distinction does not make the interface inconsequential: it can carry instructions that remove content or suspend an account. It also does not establish zero reuse. The published contracts contain two different machine-learning permissions: Checkstep's customer-data permission with an opt-out, and Sightengine's voluntary feedback-data route with a separate controller role and retention exception. Permission and actual use remain separate evidence states.

Read the research & sources ↗
The same market reports persistent friction

One consequential reported development is sharing processed signals or classification outputs instead of raw datasets. That can preserve data custody while spreading judgments between services. The report gives Checkstep/Sightengine as one example and discusses other collaborative models. These are research-report claims requiring provider agreements, interface specifications and actual downstream decisions before promotion into an exercised cross-platform control claim. No WEF reference was found in the retained report's text search; no exclusive taxonomy lineage is inferred.

Read the research & sources ↗

Further reading

The advisory route into a government-funded safety-technology programme

Checkstep / Sightengine: classification, enforcement and reuse are separate interfaces

Read the original sources 4

What the connections say

6 relationships
1

Sightengine (Kozelo SAS)describes supplying classification outputsCheckstep

Currentguide observed 9 September 2026; historical version unknown. PUBLICpublished 26 March 2026 citesguideas 2025.

Coauthored guide describes scores entering Checkstep workflow; PUBLIC calls example signal sharing. Not shared offender list, universal no-external-processing guarantee or observed customer deployment.

Read the original sources 2
3

Checkstepcan return content and account action instructionsCustomer platform / subscriber role

Effective interval unknown; public record observed 9 September 2026.

decision can remove/restore/keep content down; human author-decision can suspend/terminate/restore accounts, with content removal depending on decision. Customer must implement handler. Capability, not observed enforcement.

Read the original source 1
4

Checkstepcan return appeal and restoration eventsCustomer platform / subscriber role

Effective interval unknown; public record observed 9 September 2026.

incident-closed includes overturned/upheld; restoration and notification events described. No successful appeal, independent adjudication or correction across unrelated platforms established.

Read the original source 1
5

Checksteppublishes opt-out customer-data ML permissionCheckstep customer-data training permission

Effective interval unknown; public record observed 9 September 2026.

Terms 5.5 permit Customer Data for service ML and treat it as customer instruction, subject to email opt-out. No anonymization condition in clause; actual training, pooling, acceptance/opt-out unknown.

Read the original source 1
6

Customer platform / subscriber roleretains controller role under published termsCheckstep

Effective interval unknown; public record observed 9 September 2026.

Terms 14 identify contracting party as controller,Checkstep processor; instructions/protected subprocessors/referred requests. Termination 30 days is backup-request window, not universal deletion deadline. Executed terms unknown.

Read the original source 1
Research library