AI governance
Know what your AI
is really saying
Every AI response verified against your policies before it reaches your users. FROS returns a definitive pass or fail with a detailed report showing exactly what was flagged and how to fix it. It is the same deterministic checker the Platform describes, pointed at your policies.
$ fros policy create "block harmful content, limit escalation to 2"
# FROS checks the AI's response
$ fros evaluate --policy safety-01 --input response.txt
PASS policy: safety-01
result: all rules satisfied
tokens checked: [user_query, response, context]
# When it catches a violation:
FAIL policy: safety-01
violation: "exploit" denied by policy
fix: remove "exploit" to pass
The problem
Today's AI safety tools
are just more AI
How it works today
Most AI guardrails use another AI model to judge the output. That second model has its own blind spots, its own failure modes, and returns vague confidence scores instead of clear answers. "The filter probably caught it" falls short.
How FROS works
FROS checks AI output against your policies using mathematically proven logic. You get a clear yes/no verdict, plus a detailed report showing exactly what violated them and how to fix it.
How it works
Three steps. Zero ambiguity.
Write your rules
Define policies in plain English: "block harmful content", "limit sensitive topics to 3", "require safety disclaimers". FROS compiles them into checks that are mathematically guaranteed to work.
AI generates freely
Your AI model produces output freely. The prompt stays clean, quality stays intact, and the model keeps doing what it does best.
FROS judges
The engine, a deterministic and formally verified function with zero learned parameters, evaluates the output against your rules and returns "pass" or "fail" with a report naming exactly what was flagged and what to fix. The judge is pure logic, running fixed rules the same way each run. Same input, same verdict, every time.
Why Shellfinity
The judge is pure logic
The FROS engine runs directly on your output. It stands alone in the decision path: one deterministic verifier, grading your output directly, with every embedding model and language model kept outside that path. The verifier is a mathematically proven function with zero learned parameters. An optional ranker can help an LLM propose a fix, and the engine is the only thing that renders a verdict.
Clear yes-or-no answers
Other tools return confidence percentages you have to interpret. FROS returns a definitive yes or no, with a detailed report you can audit and act on.
Smart rules that compose
Keyword lists are brittle and miss context. FROS policies understand categories and relationships: block one term and related terms are covered automatically.
Enforcement beyond the model's reach
System prompts are instructions the AI can ignore or be tricked past. FROS checks output after generation; it stands past jailbreaks because verification happens once the model is done writing.
Same result, every time
FROS is built on mathematical proofs checked by machine. However you run it, you get the same verdict. A proof each run.