Molly — agentic threat model

Not certain from the listing — The underlying foundation models are unspecified. If vulnerable to prompt injection, attackers could manipulate Molly into executing unauthorized marketing campaigns or leaking ingested brand data.

Molly ingests public brand data, videos, spreadsheets, customer insights, and connects directly to databases. This creates a high risk of data poisoning via public channels and unauthorized access/exfiltration of sensitive customer database records.

The agent framework orchestrates campaign execution and ad optimization. Insecure tool integration with Meta/Google Ads APIs could allow an attacker to hijack tool calls, leading to unauthorized ad spend or malicious campaign deployment.

Not certain from the listing — The hosting environment and sandboxing mechanisms are undisclosed. Compromise of the SaaS infrastructure could expose highly sensitive API keys and database credentials for connected client accounts.

Not certain from the listing — While Molly claims to spot anomalies and errors in marketing data, there is no mention of security-focused observability, guardrails, or budget-limit enforcement to prevent runaway autonomous spending.

Not certain from the listing — No compliance certifications (e.g., SOC2, ISO) or identity/access management controls are detailed, raising concerns about how securely database and ad console credentials are managed.

Not certain from the listing — While Molly integrates with external ad platforms (Meta, Google, Infloso), it is unclear if it interacts with other autonomous agents or if it is susceptible to cascading failures across these ecosystems.