Open scanners vs. a managed platform.
Protect AI builds great open-source AI security tools (LLM Guard, ModelScan, NB Defense) and runs the huntr bug bounty. STACK Vault is the AI security platform you run for the team — identity, runtime, governance, compliance — without staffing the ML-inference layer yourself.
What each one is built for
STACK Vault
Managed AI security platform. Identity, runtime, agent governance, and compliance evidence — operated as one product. Multi-tenant for MSP delivery, right-sized SaaS pricing for SMB, deployment options for enterprise.
Protect AI / LLM Guard
Open-source (MIT) input/output scanners you self-host. Strong toolkit for teams that want to assemble their own runtime layer and manage the ML inference themselves. Plus the huntr platform for vulnerability disclosure.
Together
Some teams contribute back to LLM Guard while running STACK Vault for the platform layer above it — identity, agent governance, compliance evidence. Not a forced choice.
Where each one wins
Time to first protected endpoint
STACK Vault. Click, point, protected. With LLM Guard you self-host the scanner pipeline and ML inference — typically days, not minutes.
Open-source posture
Protect AI leads — LLM Guard, ModelScan, NB Defense are real open source. STACK Vault is a commercial platform; we contribute upstream where it makes sense but do not pretend to be open core.
MSP & multi-tenant delivery
STACK Vault. Tenant-isolated workspaces, per-tenant billing, MSP console. LLM Guard is per-deployment — multi-tenant orchestration is on you.
Compliance evidence at audit time
STACK Vault + Compli. Signed evidence mapped to NIST AI RMF, ISO 42001, EU AI Act, SOC 2. LLM Guard produces detection logs — turning them into auditor-ready evidence is your work.
Common questions
Is STACK Vault open source?
No. Some components publish public schemas and reference clients, but the platform is commercial. Protect AI is the right answer if open source is a hard requirement.
Can we run LLM Guard in front of STACK Vault?
Yes. Many customers do exactly this. LLM Guard handles input/output scanning at the LLM boundary; STACK Vault provides identity, audit, and compliance on top.
We are a small team. What is easiest to get started with?
For a 2-3 person team that can run Python services, LLM Guard is approachable. For a team that wants AI security as a product they consume rather than operate, STACK Vault is built for that.
Does STACK Vault have a bug bounty like huntr?
STACK Vault runs a private responsible-disclosure program with bounties. Details are on /security/.