Critical Flaws in Salesforce Agentforce Enable Zero‑Click Data Theft, Researchers Warn

Security researchers have identified three distinct vulnerabilities in Salesforce's Agentforce platform that could allow attackers to commandeer legitimate agents, siphon confidential information, and initiate phishing campaigns without any user interaction.

The flaws, collectively dubbed "SalesBleed" by the discoverers, exploit weaknesses in the way Agentforce authenticates and processes data from trusted endpoints. By leveraging a zero‑click technique, an adversary can bypass normal user prompts, gaining direct access to the internal networks of organizations that rely on the platform for customer relationship management and automated support functions.

Agentforce, a component of Salesforce's broader ecosystem, is designed to streamline interactions between support agents and customers, often integrating with third‑party tools and internal databases. The newly disclosed vulnerabilities affect the authentication handshake, data handling routines, and message routing logic, creating a pathway for malicious code to masquerade as a legitimate agent and exfiltrate data silently.

According to the original report on SecurityWeek, the attack chain begins with the exploitation of a malformed request that tricks the Agentforce server into accepting a rogue certificate. Once the forged identity is established, the attacker can issue commands that extract records, export contact lists, and even craft convincing phishing messages that appear to originate from trusted company representatives.

While no public incidents have been confirmed, the potential impact is significant for enterprises that store sensitive customer data within Salesforce. Zero‑click exploits are particularly concerning because they do not require any action from the target, such as clicking a malicious link, making detection and mitigation more challenging.

Salesforce has acknowledged the report and indicated that patches are being prepared for immediate release. In the meantime, security teams are advised to monitor for unusual Agentforce activity, enforce strict certificate validation, and apply any interim hardening measures recommended by the vendor. The disclosure underscores the ongoing need for rigorous security testing of cloud‑based services that handle high‑value data.