How to Automate Supply Chain Risk Reports: A Guide for Developers
Do you use Python? If so, this guide will help you automate supply chain risk reports using AI Chat GPT and our News API.
Two company names appear next to the same provider in a post on the Dark Web. Your correlation rule merges the two cases into one. When the next analyst opens either case, a potential case relationship is now part of the incident’s identity.
For a Dark Web API integration, a shared mention should create a relationship to review. It should not automatically merge cases.
On November 20, 2026, Bleeping Computer reported responses from JPMorgan Chase, Visa, and Mastercard about third-party incidents. Chase said its investigation suggested attackers may have accessed customers’ names and email addresses in vendors’ marketing data. Visa said its systems and payment data were not affected. Mastercard said there was limited exposure of email addresses associated with a benefits program, with no impact to the payment network.
Bleeping Computer also said that on November 16, allegedly stolen data regarding the companies was published by n0n on its extortion site. Qubole was named as the suspected provider, as opposed to a verified provider, regarding the companies. Those dates are for reporting and publishing allegedly stolen data. They do not show when the companies’ data was actually accessed.
The problem with the integration is the unresolved relationship. An attacker’s grouping of names and different company responses can be in the same research workspace without establishing one shared incident.
An API supplied by webz.io’s Dark Web service returns structured data posts from various Deep and Dark Web sources. Post documentation states that a post is identified by uuid (unique post identifier), parent post in a thread by parent_uuid and the extracted post text by text. Publication and collection times of the post are identified by respective fields.
These identifiers help the application in retaining the source material behind a suggested connection. They do not make a post a unique breach. One post can talk about multiple organizations, and several posts can be about the same allegation.
Entity enrichment can suggest names for review. Merely identifying a name in a post does not mean that the organization was breached, or that the organization used the provider mentioned in the post. It is the application’s job to bring evidence to support those claims.
Consider this constructed example. An extortion post on the deep web names Harbor Retail and Vale Logistics as Aster Analytics Customers. Your application has separate cases for both Harbor and Vale. Harbor has confirmed an exposure at an unnamed provider. Vale says they are investigating a potential 3rd Party Breach. A poorly designed rule would take the Aster label and merge the cases. It would state “Aster Incident” as an established connection. Neither of the case statements confirm this link.
Instead, add the post to both investigations as evidence of the source’s claim. Include a proposed relationship between each company and Aster, providing who made the assertion and the supporting passage. Keep the case identifiers the same.
In this design, the relationship assessment happens outside the original API record. It needs to identify the reviewer and the type of evidence that will help resolve the question. In this case, the missing evidence is confirmation of the provider involved and whether both investigations concern the same incident. A second post repeating the claim does not answer either question.
An analyst can now go to Harbor’s case, and along with the unverified Aster connection, also see the confirmed provider exposure. Vale’s case remains the same. The shared research link is visible, but it hasn’t exported Harbor’s confirmation into Vale’s case.
If a company says it’s used Aster, that would create a supplier relationship. That by itself wouldn’t indicate the event being evaluated happened at Aster. Require evidence relating the service to the incident before approving a merge.
If later evidence reveals different providers for Harbor and Vale, reject the common provider relationship and retain the reason. There is no need to restore two separate cases from a single case history, as they were never combined.
If the evidence indicates a shared incident, allow the responsible reviewer to approve that grouping. Keep the case references and the evidence justifying the grouping. The goal is to create meaningful groupings that are easily reversible, as opposed to letting a keyword match dictate a grouping.
Work through the analyst and customer views for the Harbor and Vale example. Both should show separate cases and label Aster as a source claim waiting for review. Neither should show a confirmed common-provider incident.
Now duplicate the extortion claim in the fixture. The number of source claims may vary; the claimed relationship should not be accepted just because the claim is repeated.
Lastly provide the evidence for contradictory providers in the example. Ensure that the proposed link is rejected where it is shown, and the original post is left available to show what was claimed. A line that changes status in storage but remains unqualified and solid in a customer report is still a false connection.
Do you use Python? If so, this guide will help you automate supply chain risk reports using AI Chat GPT and our News API.
Use this guide to learn how to easily automate supply chain risk reports with Chat GPT and news data.
A quick guide for developers to automate mergers and acquisitions reports with Python and AI. Learn to fetch data, analyze content, and generate reports automatically.