Numbers

Hold us to the numbers. We publish ours.

As of 16 Nov 2026

01Side by side

Our figures beside the ones vendors publish. Each with its source.

Attacks caughtHigher is better
  1. Kriton ShieldKriton100%14 of 14 attack prompts, every one of 5 runs. Run file on request
  2. LLM GuardProtect AI99.7%Recall on 20,000 prompts. Source
  3. NeMo GuardrailsNVIDIA98.9%With three safety models, 215 in-house cases. Source
  4. Lakera GuardCheck Point95.2%Its own PINT benchmark. Source

Not published: Prompt Shields, Bedrock Guardrails, Model Armor, Lasso for Claude Code, Prisma AIRS for Claude Code, Snyk Studio hooks, AI Guard for Claude Code. Each vendor tests on its own set, so read these as published claims side by side.

Time added to a promptLower is better
  1. Lakera GuardCheck PointUnder 40 msShort prompts, setup not published. Source
  2. LLM GuardProtect AI104 msOn a CPU; 7.65 ms on a GPU. Source
  3. Kriton ShieldKriton341 msMedian of 210 decisions, our side only. Run file on request
  4. NeMo GuardrailsNVIDIA530 msAverage added with three safety models. Source

Not published: Prompt Shields, Bedrock Guardrails, Model Armor, Lasso for Claude Code, Prisma AIRS for Claude Code, Snyk Studio hooks, AI Guard for Claude Code. Ours is the decision on our side, before network time. Vendors measured on their own setups.

Ordinary prompts blockedLower is better
  1. Lakera GuardCheck PointUnder 0.5%Calibrated production setups. Source
  2. Kriton ShieldKriton6%6 of 100; 5 are one prompt with the word password. Run file on request

Not published: Prompt Shields, Bedrock Guardrails, Model Armor, NeMo Guardrails with NemoGuard NIMs, LLM Guard, Lasso for Claude Code, Prisma AIRS for Claude Code, Snyk Studio hooks, AI Guard for Claude Code. We publish ours even where it is not the best number on the chart.

02How we measured

Our measured numbers

Attack prompts stopped
14 of 14
Every run, 5 runs of 14 attack prompts, as shipped, 8 Nov 2026. Run file on request.
Prompt decision, server side
341 ms
Median of 210 decisions (42 prompts, 5 runs each); 582 ms p95; none over 3 s; 8 Nov 2026. Time on our side, before network and start-up on the person’s machine. Run file on request.
Ordinary prompts blocked
6 of 100
20 ordinary prompts, 5 runs each, as shipped; 5 of the 6 are one ordinary prompt with the word password in it; 8 Nov 2026. Run file on request.
When Kriton cannot answer
Your setting decides
You choose to let short prompts through or stop them. 0 errors in 210 decisions, 8 Nov 2026. Run file on request.
03How we compare

Before the action, or after it. That decides what a guard can stop.

When each guard acts, whether it works in Claude Code, and whether its vendor publishes latency, detection and false-block figures
ProductWhen it actsIn Claude CodeLatencyDetectionFalse blocks
Kriton ShieldKriton Agentic SecurityOn the prompt before the agent sees it, and before each agent action runs Yes Published, above Published, above Published, above
Lakera GuardCheck Point AI Guardrails (Lakera) When your code calls the API: before the model call, after the response, or both, including tool callsSource, 16 Nov 2026 No published integration; API call from your codeSource, 16 Nov 2026 PublishedSource, 20 Oct 2025 PublishedSource, 2 May 2025 PublishedSource, 16 Nov 2026
NeMo Guardrails with NemoGuard NIMsNVIDIA Input, dialog, execution and output rails, tools includedSource, 16 Nov 2026 No published integrationSource, 16 Nov 2026 PublishedSource, 3 Mar 2025 PublishedSource, 3 Mar 2025 Not published by the vendorSource, 16 Nov 2026
LLM Guard, prompt injection scannerProtect AI (Palo Alto Networks) On the prompt, wherever your code runs the scannerSource, 19 May 2025 No published integrationSource, 16 Nov 2026 PublishedSource, 19 May 2025 PublishedSource, 16 Nov 2026 Not published by the vendorSource, 16 Nov 2026
Lasso for Claude CodeLasso Security Enterprise: on the request and before each tool call. Open-source hook: after the tool runs, warns without blockingSource, enterprise, 16 Nov 2026Source, open source, 16 Nov 2026 Yes, via Claude Code hooksSource, 16 Nov 2026 Not published by the vendorSource, 16 Nov 2026 Not published by the vendorSource, 16 Nov 2026 Not published by the vendorSource, 16 Nov 2026
Prisma AIRS for Claude CodePalo Alto Networks Before the model processes a prompt or tool call, and on responses and tool outputsSource, 19 Nov 2025 Yes, via Claude Code hooksSource, 19 Nov 2025 Not published by the vendorSource, 16 Nov 2026 Not published by the vendorSource, 16 Nov 2026 Not published by the vendorSource, 16 Nov 2026
AI Guard for Claude CodeZscaler On the prompt, before MCP and web tools run, and on their responsesSource, 16 Nov 2026 Yes, via UserPromptSubmit, PreToolUse and PostToolUse hooksSource, 16 Nov 2026 Not published by the vendorSource, 16 Nov 2026 Not published by the vendorSource, 16 Nov 2026 Not published by the vendorSource, 16 Nov 2026
04Why before matters

247 days to find and contain a breach. Kriton stops the action before it runs.

247 days (183 to identify, 64 to contain), up from 241. IBM, Cost of a Data Breach Report 2026, published 10 Aug 2026.

Bring your own prompts. We run the bench live.