Verify LLM Output

Block AI responses that violate regulatory or corporate policies before they reach users. Formal verification using Z3 theorem proving.

$ pip install aare-core

view source

guardrail.py

from aare import HIPAAGuardrail
from langchain_openai import ChatOpenAI

llm = ChatOpenAI()
guardrail = HIPAAGuardrail()

# verify output before it reaches users
chain = prompt | llm | guardrail

try:
    response = chain.invoke({"text": user_input})
except ViolationError as e:
    # blocked - policy violated
    log_violation(e.result)

// pipeline

input

User Query

generate

LLM Response

verify
Aare

output

Safe Response

Current guardrails fail

Prompt engineering: "please don't violate policies"
Regex filters: brittle, easy to bypass
Human review: doesn't scale
Trust the model: jailbreaks happen

Aare provides proof

Formal verification via Z3 theorem prover
Post-generation: immune to prompt injection
Audit-ready proof traces
Configurable: block, warn, or redact

// use cases

live demo

HIPAA Compliance

Block PHI from AI responses in healthcare apps. All 18 Safe Harbor categories.

PCI DSS

Prevent credit card numbers and cardholder data in responses.

Corporate Policy

No competitor mentions, pricing commitments, or legal advice.

Content Safety

Block harmful content before it reaches users.

Data Leakage

Prevent internal data and API keys from being exposed.

Legal Compliance

GDPR, CCPA, and jurisdiction-specific regulations.

// hipaa demo

See verification in action

llm output

result

// click verify to check for violations

api log

[+]

// waiting for api call...