What is an AI guardrails platform?

An AI guardrails platform enforces policy on AI inputs and outputs at runtime, blocking harmful or non-compliant content before it reaches a user. SUPERWISE evaluates PII, toxicity, topic, and jailbreak policy on every request in under 10ms, as part of the AMP control plane.

Do guardrails run at runtime or in review?

At runtime. Each request is checked before the response is returned, so a violation is blocked as it happens rather than found in a later review. The block is logged, so you have the evidence too.

Can I define my own policy rules?

Yes. Alongside built-in checks for PII, toxicity, topic, and jailbreak, you can add custom policy rules for your own restricted topics, brands, and content, and apply them across every model and agent.

How do guardrails fit with the rest of AI governance?

Guardrails are the runtime layer of SUPERWISE AMP, the AI governance platform. The same policy that blocks a request produces the audit trail, so enforcement and evidence come from one system.

SUPERWISE® AMP → Guardrails

The AI guardrails platform that blocks failures before they reach users

Enterprise AI guardrails are runtime product controls that evaluate live AI requests and enforce policy before a response reaches a user.Framework guardrails define expectations, and code guardrails protect one application path. SUPERWISE runtime checks apply policy consistently across governed AI traffic.

<10ms latency99.99% uptime

Start building free Talk to us

What the AI guardrails platform does

Guardrails are the runtime layer of the SUPERWISE AI governance platform. Every request and response is checked against your policy for PII, toxicity, restricted topics, and jailbreak attempts, and a violation is blocked before the user sees it.

The same policy covers the agents you build on the platform, SUPERWISE Chat, and the AI tools your team already uses through Sentinel, the AI gateway, which redacts PII before a prompt reaches a public LLM. Every block lands in one audit trail.

<10ms

Policy Evaluation

10B+

Inferences Protected

99.9%

Detection Accuracy

Data Breaches

See It Work

Guardrails in Action

superwise-guardrail-monitor

● Monitoring active...

Real-time policy enforcement in < 10ms

Built-in Protections

Pre-configured guardrails for common risks, plus the flexibility to define your own

99.9% accuracy

PII Detection

Patient SSN in the response. Caught. Redacted. Zero leaks.

<5ms latency

Toxicity Filter

Your agent just insulted a customer. Prevent harmful responses in real time.

Zero false negatives

Topic Restriction

User tried to make your agent discuss competitors. Topic blocked.

95% detection rate

Jailbreak Detection

Attacker tried to override system prompt. Attempt logged and rejected.

Multi-layer defense

Competitor Check

Agent almost named a competitor. Blocked before the word left.

Protection at Every Layer

Input Validation

Detect prompt injection, PII, and malicious inputs before model execution.

Runtime Evaluation

Evaluate AI outputs against your policies in real-time. Block, modify, or flag violations automatically.

Continuous Learning

Guardrails improve over time based on your feedback and new threat patterns.

# Apply guardrails to any LLM call
response = sw.chat.complete(
  model="gpt-4",
  messages=messages,
  guardrails=[
    "pii",           # Block PII
    "toxicity",      # Content safety
    "data_leakage",  # Prevent leaks
    "hallucination", # Fact checking
    {
      "type": "custom",
      "policy": "no_competitor_mentions"
    }
  ]
)

# Guardrail results
print(response.guardrails)
# {
#   "passed": true,
#   "checks": [...],
#   "latency_ms": 8
# }

Guardrails Stop Threats in Real-Time

Watch SUPERWISE detect and block PII leaks before they ever reach the LLM. See how a patient SSN gets stopped cold.

Talk to us Read the docs

Related Capabilities

As part of SUPERWISE AMP, guardrails work together with other capabilities

AI guardrails platform questions

What buyers ask when comparing AI guardrails platforms

Questions buyers ask

All questions

Deploy Your First Guardrail Free

Minutes, not weeks. Stop your next breach before it happens.

Talk to us Watch It Work in 60 Seconds