Salesforce AI Hack Enabled CRM Data Theft

Salesforce’s Agentforce platform is reportedly vulnerable to a threat described as the ForcedLeak method, which researchers say could be used with prompt injection and an expired domain to enable CRM data theft. The technique was identified by Noma Security, a company recently behind AI security revelations. Agentforce allows businesses to build and deploy autonomous AI agents across sales, marketing and commerce, with these agents able to complete multi-step tasks without constant human intervention.

The ForcedLeak method centres on Agentforce’s Web-to-Lead functionality, which creates a web form for external users to submit lead information that is saved into the CRM. The findings were discussed in an article according to Security Week. This article notes that the attack could potentially enable unauthorised access to CRM data via a standard data capture form.