Plannix Bot security FAQ
Explore detailed answers to security, compliance, data sovereignty, and isolation questions frequently raised by enterprise risk management and CISO teams.
Is our data used to train models?
Strictly No. Under our Terms of Service and Enterprise Master Service Agreements (MSA), Plannix commits never to train or fine-tune public models using your prompts, source code, artifacts, or execution traces. Your data remains your exclusive intellectual property.
Where is code stored during execution?
Source code is cloned onto an ephemeral virtual drive encrypted with a per-session AES-256 key. Ephemeral drives are destroyed upon sandbox shutdown and their blocks are zero-filled according to DoD 5220.22-M standards.
How are container breakouts prevented?
Rather than standard unprivileged Docker containers sharing a host Linux kernel, Plannix utilizes hardware microVMs (KVM/Firecracker). Even if arbitrary malicious shellcode achieved root privileges inside the sandbox, it cannot escape the hardware hypervisor boundary.
Which LLM backends process our code?
Plannix routes requests to dedicated private model endpoints hosted in SOC 2 Type II compliant datacenters. Direct VPC endpoints can also be configured to route to self-hosted models or private Azure OpenAI / AWS Bedrock VPC deployments.
Which compliance certifications do you maintain?
Plannix maintains SOC 2 Type II certification, ISO 27001, ISO 27701, GDPR compliance, and can execute Business Associate Agreements (BAAs) for HIPAA-regulated environments. Audit reports are available upon request through our Trust Center.