Detect Deepfakesby Resemble AI
Agentic AI attack

Salesforce customer environments agentic AI attack

Salesforce reports threat actors using the ARTEX agentic framework to conduct automated, large-scale attacks against customer environments via credential.

Reported date
Nov 12, 2026
Target
Salesforce customer environments
Agent type
Other AI agent
Agent role
Used by the attacker

The exact incident date was not established. This entry is dated by its source report.

Updated Nov 19, 2026 · 1 min read

What happened

Salesforce Threat Intelligence has identified threat actors utilizing the ARTEX framework to conduct automated, large-scale attacks against Salesforce customer environments. ARTEX is an open-source, multi-agent offensive framework designed for autonomous penetration testing, which allows attackers to provide a target and objective for AI agents to coordinate intrusions. While the original repository was removed, community forks remain in circulation, enabling attackers to perform continuous planning, parallel execution, and on-the-fly code generation.

In the observed campaigns, attackers leveraged stolen application credentials and trusted integration sessions to gain initial access. Salesforce clarified that this activity did not stem from a platform vulnerability. Once inside, the agentic framework demonstrated capabilities that exceeded standard human operator speeds, including the rapid, automated deployment of custom Apex code tailored to the specific environment.

Key tactics observed include:

  • Environment Relaying: Using Apex and abused Named Credentials to reach connected services and facilitate lateral movement.
  • Custom Code Deployment: Inspecting existing Apex source and integration settings to deploy custom REST endpoints and Visualforce components for persistent access.
  • Permission Enumeration: Testing the effective permissions of compromised accounts by enumerating users, profiles, and permission sets to attempt privileged operations.
  • Adaptive Execution: Automatically iterating through multiple code-execution and integration-access paths, including anonymous Apex and various authentication configurations, to recover from unsuccessful attempts.

Salesforce noted that the operational scale and velocity of these attacks strongly indicate the use of agentic operators. The framework allows malicious actors to sustain multiple operations simultaneously and navigate complex environments at a speed that challenges traditional human-led incident response. Salesforce has directly notified affected customers and recommends implementing phishing-resistant MFA, enforcing IP allowlisting, and utilizing behavioral detection to identify complete attack sequences rather than isolated events.

Evidence in the reporting

Incident evidence
We have observed a stolen-credential-driven campaign targeting Salesforce customer environments
Agent involvement
audit logs show the activity cluster rapidly and dynamically writing automation

Sources