Is my AI tool safe?Assistants and personal agentsSalesforce
Are Salesforce Agentforce and connected agents safe?
No tool is safe or unsafe on its own: it depends on what it can reach and how it is set up. The record for Salesforce Agentforce and connected agents: 2 incidents in this center's library (its advisory feed does not track it, so it has no CVE count here). Below: what went wrong, what was fixed, and how to use it safely.
What to watch for
ForcedLeak turned a web form into a channel out of the CRM, and the Salesloft Drift breach used stolen agent tokens to raid Salesforce data at hundreds of companies. The agent's tokens are worth more than any one user's.
How to use it safely
Inventory every connected app and agent token, scope and rotate them, and treat customer-submitted text as attacker-controlled input to agents.
What it can reach
- Reads attacker-controlled content
- Connected to mail, files, chat or CRM
- Holds API keys and tokens
What went wrong: incidents in the library
- ForcedLeak: Salesforce Agentforce leaks CRM data through a web form Critical
An attacker could type hidden instructions into a company's public lead form. When an employee later asked Agentforce to handle the lead, the AI sent CRM data to the attacker through an expired domain that Salesforce still… - Salesloft Drift: stolen AI chat agent tokens used to raid Salesforce data Critical
Attackers stole the OAuth tokens that the Drift AI chat agent used to connect to customers' Salesforce accounts, then bulk-exported CRM data from hundreds of companies, including security vendors, and hunted for passwords and…
A high CVE count can mean heavy scrutiny as much as weak engineering, so read the incidents and advisories, not only the number. The counts are what this center has recorded, not every flaw ever found; usage figures are the vendors' and registries' own published numbers.