
09/30/2025
Salesforce AI Hack Exposes a Bigger Problem!
Noma Security’s discovery of ForcedLeak, a prompt injection attack that targeted Salesforce’s Agentforce platform, shows just how quickly attackers are adapting to exploit AI-powered systems.
Here’s what happened:
- A malicious payload slipped into Salesforce’s Web-to-Lead forms could trick AI agents into exfiltrating sensitive CRM data.
- An expired, trusted Salesforce domain left the door open for attackers to quietly receive that data.
-The attack had a high chance of going undetected.
This incident is a warning: as enterprises integrate AI agents into critical workflows, they are also multiplying their attack surface. CRM data, customer trust, and brand reputation all hang in the balance.
At KATSH.ID, we see the lesson clearly: identity and access management must evolve in sync with AI adoption. Securing the digital economy requires trusted, fraud-resistant digital identities that keep both people and machines in check.
The AI-driven economy will only thrive if digital identities are trusted, resilient, and fraud-resistant. That’s the future KATSH.ID is building.
Source: https://www.securityweek.com/salesforce-ai-hack-enabled-crm-data-theft/