Lakera alternatives: the 2026 comparison

Honest answer: for many enterprises, it still is. Lakera has the strongest brand in the category, adversarial data collected at scale through the Gandalf game — Check Point's announcement cites over 80 million adversarial patterns — a published prompt-injection benchmark (PINT), and a documented self-host path including fully air-gapped deployments. If you are already a Check Point customer, or you want the category leader with enterprise support behind it, staying with (or choosing) Lakera is a defensible decision. The trade-offs are the classic post-acquisition ones: sales-gated pricing and a roadmap that now serves the Check Point platform first.

LLM Guard was the reference open-source option: an MIT-licensed Python library with 15 input and 20 output scanners you compose into your own pipeline. Nothing leaves your infrastructure and it costs nothing. The caveat is now decisive: the repository was archived on July 9, 2026 — a year after Palo Alto Networks completed its acquisition of Protect AI — and its README states the project and its models are no longer maintained. Choosing it today means forking and maintaining a frozen detection stack yourself, against attacks that evolve monthly. If you want open source that is still moving, look at NeuralTrust's TrustGate instead.

If your stack already runs on Cloudflare and your primary need is operational — usage analytics, caching, rate limiting, retries and model fallback across providers — AI Gateway is hard to beat: the core features are free on any Cloudflare account. Be precise about what it is not, though. Its Guardrails feature is content moderation (harmful-content categories, powered by Llama Guard); prompt-injection detection lives in a separate product, Cloudflare's AI security add-on in the WAF, gated to Enterprise plans. There is no self-hosting, and AI Gateway offers no regional processing controls, which matters if EU data residency is a requirement.

NeuralTrust (Barcelona) is the other independent European option in this list. Its TrustGate gateway is open source (Apache-2.0, written in Go), self-hostable via Docker, Kubernetes or a single binary, and the platform adds runtime protection, posture management and red teaming (TrustTest). The company raised a €17.2M seed in June 2026 and is independent as of this writing. The honest counterpoint: community adoption of the open-source gateway is still small, so expect to lean on the vendor rather than a large ecosystem.

HiddenLayer is not primarily a prompt firewall — it is a broad AI-security (MLSecOps) platform: scanning model files for backdoors and serialization exploits across 35+ formats, AI asset discovery, attack simulation, plus a runtime detection module. If your risk register is dominated by model supply chain — third-party model files, fine-tunes, registries — HiddenLayer (or Palo Alto's Prisma AIRS) addresses problems a runtime proxy like Lakera or Senthex does not. Pricing is sales-led with no public numbers, and the deployment model is not publicly documented, so plan a real procurement cycle.

Disclosure: Senthex is us — judge this section accordingly, and check the claims against the linked pages.

Senthex is an independent, European runtime LLM firewall: a proxy between your app and OpenAI, Anthropic, Mistral or Gemini, integrated by swapping a base URL — one line of code. It is EU-hosted (Hetzner Falkenstein, Germany) and self-hostable, including on Azure. Raw request and response bodies are excluded from storage by default (data minimisation). Two things are genuinely distinctive. First, the audit trail is designed to be verifiable: in the v1.1.7 pilot, each record is chained with SHA-256 and anchored to an RFC 3161 timestamp so you can verify it offline — tamper-evident, not tamper-proof — aligned with EU AI Act record-keeping (Articles 12 & 19). You can verify a real bundle yourself. Second, we publish original research on multi-agent attacks — the ATLAS and RELAY studies — with frozen datasets, so you can see how the firewall behaves before buying. Pricing is public, with a free tier.

When Senthex is not the right choice: if you need model-file scanning and AI supply-chain security (HiddenLayer, Prisma AIRS); if you are all-in on Cloudflare and only need gateway operations (AI Gateway); if you want a large-scale enterprise red-teaming program attached to the category leader (Lakera/Check Point); or if your policy is to run only self-maintained open source (TrustGate). See also what we do and do not claim on EU AI Act compliance.