Lakera Guard alternatives: 10 prompt-injection guardrails compared

Last updated October 26, 2026. Each listing links to its full page with sources. No lab results yet: this is a fact comparison, not a ranking by test score.

Short answer: Lakera Guard still exists. Check Point closed its acquisition of Lakera on October 22, 2025, and the product is now sold as Check Point AI Guardrails. If you want an alternative, the right one depends on where you run your models:

• You want open source you can self-host: NeMo Guardrails (NVIDIA, Apache-2.0), Guardrails AI (Apache-2.0), or Meta's PurpleLlama tools (LlamaFirewall is MIT-licensed).
• You're already on a big cloud: Amazon Bedrock Guardrails, Azure AI Content Safety Prompt Shields, or Google Cloud Model Armor. Bedrock (through its ApplyGuardrail API) and Model Armor (through its REST API) can screen content from models outside their own platform, per their own docs.
• You already use Cloudflare's WAF: Cloudflare AI Security for Apps (formerly Firewall for AI) adds prompt-injection detection at the edge.
• You need an enterprise platform that also covers employee AI use: Prompt Security (SentinelOne).
• You want a gateway with self-serve pricing: NeuralTrust (free tier, Standard $49/month).
• You run Fortinet: Fortinet FortiAIGate.

What Lakera Guard does today

Check Point's docs describe the Guard API as screening prompts, outputs, tool calls, tool responses and tool descriptions for prompt attacks, data leakage, content violations, malicious links and off-policy agent behavior. The higher AI Agent Security tier adds agent discovery and risk assessment. No price list is published; the docs say you can sign up for a free account to start. Full listing

Comparison table

Open-source alternatives

NeMo Guardrails (NVIDIA). An open-source toolkit for adding programmable guardrails (topic, safety and jailbreak rails) to LLM apps. Apache-2.0, per the repo's LICENSE file. Pick it if you want to define conversation rules in code and host everything yourself. Listing

Guardrails AI. An open-source framework and validator hub for checking and correcting LLM inputs and outputs, Apache-2.0. Harvey announced its acquisition of Guardrails AI on September 8, 2026. The open-source repo was still active on October 25, 2026. Pick it if you want pluggable validators in a Python app. Listing

PurpleLlama (Meta). Meta's set of open tools to assess and improve LLM security, including Llama Guard classifiers, Prompt Guard and LlamaFirewall. The licenses are mixed: evals, Code Shield and LlamaFirewall use MIT, while the models use the matching Llama Community License. Check that license before you ship the models commercially. Listing

Cloud-native alternatives

Amazon Bedrock Guardrails. Content and word filters, denied topics, PII redaction, prompt-attack detection, plus contextual grounding and Automated Reasoning checks for hallucinations. Through the ApplyGuardrail API it can screen any model, not just Bedrock models. Usage-based AWS pricing. Listing

Azure AI Content Safety Prompt Shields. Detects user prompt attacks (formerly called jailbreak risk detection) and document attacks, meaning indirect prompt injection hidden in documents, emails or web pages. For Foundry agents it can also scan tool responses for supported tools (preview). Usage-based Azure pricing. Listing

Google Cloud Model Armor. Screens prompts, responses and agent interactions for prompt injection, jailbreaks, sensitive data leaks, malicious URLs and harmful content. It works with any model through a REST API. Google publishes the price on the product page: free up to 2 million tokens a month, then $0.10 per additional 1 million tokens. Listing

Cloudflare AI Security for Apps (formerly Firewall for AI). A WAF detection that flags prompt injection, PII and unsafe or custom topics in requests to endpoints labeled cf-llm. LLM endpoint discovery is on every plan, but the detection fields are an Enterprise paid add-on. Listing

Commercial platforms

Prompt Security (SentinelOne). Covers employee AI use, AI code assistants, homegrown AI apps and agents, with shadow-AI discovery, data-leak prevention, prompt-injection and jailbreak defense, and agent and MCP discovery. It's built into SentinelOne's Singularity Platform. Listing

NeuralTrust. An AI gateway (TrustGate) and guardrails (TrustGuard). It's one of the few vendors here with self-serve pricing: a free plan with 100,000 requests a month, and Standard at $49 a month as usage credit. Listing

Fortinet FortiAIGate. A self-hosted AI security gateway whose guardrails block prompt injection, jailbreaks and data leaks in prompts, outputs and MCP tool calls, with per-team budgets and rate limits. Fortinet also announced it on October 26, 2026 as the cloud-delivered FortiGuard Inline AI Security Service for FortiGate customers. Listing

Be careful with these

• LLM Guard (Protect AI, now part of Palo Alto Networks): the GitHub repo was archived on July 9, 2026, and its README says it is no longer developed or maintained. Listing
• Vigil: Apache-2.0, but the last code push was in January 2024. Listing

How to choose

1. Where does your traffic run? If everything is on one cloud, start with that cloud's guardrail. You'll get one bill and one console.
2. Do you need agent coverage, not just chat? Look for tool-call and tool-response screening. Check Point AI Guardrails, Azure Prompt Shields (preview, for Foundry agents), Model Armor (agent interactions) and FortiAIGate (MCP tool calls) say so in their own docs.
3. Can you self-host? NeMo Guardrails, Guardrails AI and LlamaFirewall run in your own environment. FortiAIGate is self-hosted too.
4. Is the price public? In this category, 8 of the 13 tools that are not purely open source have no public price on record in our data (as of October 26, 2026). The cloud services bill by usage, Google Cloud Model Armor lists exact per-token prices, and NeuralTrust has self-serve plans.

See every guardrail we track in Prompt-injection firewalls & guardrails, or download the open data.

Sources checked on October 26, 2026

• Lakera Guard: https://docs.lakera.ai/guard.md and Check Point's Q3 2025 results release, https://www.checkpoint.com/press-releases/check-point-software-reports-2025-third-quarter-financial-results/
• NeMo Guardrails: https://github.com/NVIDIA-NeMo/Guardrails (LICENSE.md: Apache-2.0)
• Guardrails AI: https://github.com/guardrails-ai/guardrails and https://guardrailsai.com/blog/guardrails-ai-joins-harvey
• PurpleLlama: https://github.com/meta-llama/PurpleLlama (README license table)
• Amazon Bedrock Guardrails: https://aws.amazon.com/bedrock/guardrails/ and https://aws.amazon.com/bedrock/pricing/
• Azure Prompt Shields: https://learn.microsoft.com/en-us/azure/ai-services/content-safety/concepts/jailbreak-detection and https://azure.microsoft.com/en-us/pricing/details/cognitive-services/content-safety/
• Google Cloud Model Armor: https://cloud.google.com/security/products/model-armor
• Cloudflare AI Security for Apps: https://developers.cloudflare.com/waf/detections/ai-security-for-apps/
• Prompt Security: https://www.sentinelone.com/platform/securing-ai-prompt/ and the SEC filing at https://www.sec.gov/Archives/edgar/data/1583708/000110465925088079/tm2525181d1_8k.htm
• NeuralTrust: https://neuraltrust.ai/pricing
• Fortinet FortiAIGate: https://www.fortinet.com/products/fortiaigate
• LLM Guard: https://github.com/protectai/llm-guard (archived); Vigil: https://github.com/deadbits/vigil-llm

All guardrail tools Open data (JSON) Methodology