Our takeThree flaws in Salesforce Agentforce let attackers hijack trusted agents for zero-click CRM data theft and anonymous phishing. Patch and review your Agentforce configurations now.Cyber Resilience desk
Sources (3)
- register_sec · register_sec
- securityweek · securityweek
- infosec_mag · infosec_mag
What this means for you — Security leader:Patch Agentforce immediately and review all agent permissions and prompt boundaries. Disable or restrict any public-facing agent endpoints that could be hijacked for data exfiltration or phishing.
What this means for you — Lean IT orgs:If your team uses Salesforce Agentforce, apply the patches right away and turn off any public internet access to it. Check that only the people who need it can reach your CRM data.
What this means for you — MSP:Audit every client using Salesforce Agentforce for the three SalesBleed CVEs, push the patches, and verify that agent endpoints are not internet-exposed. Review agent permissions and add monitoring for anomalous data access or outbound phishing attempts.
What this means for you — Researcher:The three Agentforce flaws demonstrate how prompt-injection-style hijacks in trusted agents can lead to zero-click data theft and anonymous outbound phishing. Worth testing similar agent boundaries in other CRM and automation platforms.