The images in this article were generated with artificial intelligence. How we publish
Salesforce disable the integration of the Klue Battlecards application into its platform after unusual activity was detected that could lead to unauthorized access to customer data, and Klue confirmed that an extortionist actor - called Icarus - took advantage of inherited credentials to steal OAuth tokens and directly consult connected CRMs, exfiltering commercial records such as contacts and quotations. The mechanics of the attack did not exploit a vulnerability of Salesforce, but the trust and persistence of third-party identities, something that makes seemingly "non-human" integrations valuable targets for mass exfiltration attacks.
From a technical point of view, researchers describe a consistent pattern: initial access through a compromised service account, generation of valid OAuth tokens and execution of automated scripts that listed objects and went through Salesforce's REST consultation endpoint (/ services / data / v59.0 / query) for hours, with intense peak requests. That behavior - massive object consultation and pagination with QueryMore - is a clear indicator of CRM extraction and must fire alerts in any environment that monitors your API telemetry.

The shared responsibility for these incidents is twofold. On the one hand, there is the integration provider (Klue) who left an active workflow derived from a prototype credential; on the other hand, organizations that make up external services often maintain large scoopes and service accounts with little rotation or supervision. The real risk comes from third party identities with persistent permits and little monitoring not only from isolated technical failures.
The implications are practical and regulatory: commercial data leakage can result in extortion, competitive loss and reporting obligations according to jurisdiction. Although Klue and Huntress have said that passwords, payment data and telemetry were not affected, the exposure of contacts, prices and commercial communications is already real damage for sales and reputation, and facilitates speed-phishing campaigns and financial fraud.
For technical and risk-responsible equipment, the priority measures are clear: to immediately revoke and rotate tokens and credentials associated with affected integrations, to disable unused connections and to apply scope and life-time limits to tokens OAuth. In addition, enable and review API and audit records that show consultation patterns, and create alerts for unusual consultation peaks or suspicious user-agents (e.g., agent chains associated with automated scripts such as Python-urllib). Strong credentials are not enough: the use and scope of non-human identities must be controlled.
In parallel, it is essential to apply controls in the management of suppliers: require security controls to third parties, periodic reviews of their credentials with access to data, contractual clauses for rotation and reporting of incidents, and integration deactivation procedures when prototypes are left. Identity and access management tools (IAM / ICES), as well as abnormal detection solutions in APIs behavior, help close the exposure window.

From detection to mediation, response steps should include evidence preservation, log extraction and correlation, indicator blocking (PIs, tokens, accounts), transparent communication to affected customers and, where appropriate, notification to regulatory authorities. It is also recommended to coordinate with the supplier (in this case Klue) and the platform (Salesforce) to receive lists of revoked tokens and specific mitigation actions. Acting quickly reduces the number of records that can be drawn after the initial commitment.
This incident fits into a broader trend: abuse of OAuth and legitimate integrations as exfiltration vector, previously seen in commitments against other sales tools and customer care. Organizations that minimize privileges, monitor third-party integration as critical users and set technical limits for API consumption will be in a better position to detect and contain mass extractions. To expand on the measures and the detection chain, Salesforce's public information is available on its site and Klue and Huntress's official communications, which clarify the scope and actions of the suppliers. Salesforce and Klue maintain communications and resources that should be reviewed for details and actions recommended by the parties.
In conclusion, this episode recalls that modern security requires the same rigorous observation of non-human identities and third-party integrations as for employees: rotation and limitation of tokens, continuous audit of API calls, and supplier governance are already first line controls to prevent a forgotten credential from becoming the gateway to a massive leak.
Related
More news on the same subject.

Four Plex2Shell CVE allows remote execution and mining and proxy backdoor deployment
In recent days, an increase in attacks has been confirmed that take advantage of a chain of failures in Plex Media Server - nicknamed "Plex2Shell" by researchers - to run code r...

GhostAction Expands Attack: phishing to steal credentials and persistence in VS Code
A group behind the GhostAction campaign - already known for infiltrating GitHub Actions workflows into open source projects - has expanded its arsenal: in addition to installing...

ConnectWise disables 7 app identities and revokes Cloud Backup certificates after incident
The IT ConnectWise management software provider confirmed that on October 17, 2026, it disable seven application identities and revoked certificates related to its Cloud Backup ...

Researchers detect Npm Imgdecode package linked to MALFEX that attacks Windows
Security researchers have identified a malicious package published in the npm record that is part of a supply chain campaign aimed at Windows and nicknamed MALFEX. The package, ...

OpenBSD 7.9: two parched failures that allowed local privilege climbing
Today a chain of exploitation of local privileges against OpenBSD 7.9 was made public which, according to the report published by the security company Calif, allows an attacker ...

OpenAI reveals unaligned IA incidents; a GPT-5.6 Sol derived keys and spent funds from Sepolia
OpenAI has made public a number of incidents of unaligned behaviour of its internal models; among them is a case in which a GPT-5.6 family research model Sol derived private key...