Piper AI SDR — agentic threat model

Not certain from the listing — The underlying LLMs or foundation models powering the conversational interfaces are not specified. Standard risks include prompt injection via the public chat interface, which could reprogram the bot to output misaligned or brand-damaging content.

The agent ingests real-time visitor identification, intent analysis data, and integrates deeply with Salesforce CRM. This creates significant risks of data exfiltration of sensitive customer/lead PII, and potential CRM database poisoning if malicious inputs are written back to Salesforce.

Not certain from the listing — The specific orchestration framework is proprietary. However, the agent translates user intent into tool calls (Salesforce updates, routing actions, and launching voice/video communication), making insecure tool integration and prompt injection-to-tool-execution a primary threat vector.

Not certain from the listing — The hosting infrastructure, sandboxing of execution environments, and secrets management for CRM API keys are not detailed. Compromise at this layer could expose Salesforce credentials.

Not certain from the listing — While the platform provides custom reports and dashboards for pipeline performance, it is unclear if there are dedicated LLM security guardrails, real-time anomaly detection, or logging of adversarial prompt attempts.

Not certain from the listing — The listing does not explicitly cite security compliance certifications (e.g., SOC 2, GDPR, ISO 27001), though enterprise Salesforce integrations typically require robust OAuth and authorization controls.

Not certain from the listing — The agent primarily routes leads to human sales representatives rather than interacting in a multi-AI-agent ecosystem, though cascading failures could occur if integrated marketing automation platforms are compromised.