Safeguard

What Safeguard does

Safeguard.sh Inc, founded in 2024 and headquartered in Dublin, California, is an AI-native software supply chain security platform that combines proprietary purpose-built models with autonomous remediation. The company's core technology is a model family — Griffin (five variants from 8B to 671B-MoE), Eagle (13B), and Lion (1B distilled) — trained exclusively on an 11M+ document cybersecurity corpus using a security-augmented tokenizer and an adversarial disproof decoder head, served on the Aegis architecture. The platform delivers 100-layer call-graph reachability analysis (40 layers deeper than legacy SCA), 11 fused scanners across code, containers, and IaC, and Auto-Fix that authors, tests, and merges remediation pull requests under policy gates. Its product surface includes ESSCM, SBOM Studio, TPRM, Open Source Manager, IaC Security, DAST, Safeguard Guard (runtime LLM protection), Cowork, and a Gold Registry of 500K+ zero-CVE hardened components, with 373 compliance framework mappings including FedRAMP HIGH, IL7, SOC 2, ISO 27001, NIST SSDF, EO 14028, DORA, NIS2, and EU AI Act.

The business model is usage-based pricing positioned at roughly a tenth of per-seat Snyk-class ($1,200/dev/yr) and below Vanta-class GRC pricing, with consumption tiers spanning shared cloud, dedicated clusters, VPC-isolated, and sovereign/air-gapped deployments (multi-year contracts) running Griffin Zero on configured 11x H100 to 22x H100 multi-AZ footprints. Revenue also flows from enterprise support and services, plus a Marketplace with 40 community-authored policies and partner-built scanner SDK. GTM is multi-modal: a free 7-day sandbox with self-serve conversion, direct enterprise field sales targeting regulated industries (financial services, banking, defense, healthcare, government), and an MCP-native distribution channel where AI agents (Claude, Cursor, Cline) can evaluate, subscribe, and pay end-to-end over Model Context Protocol. Geographic reach spans nine US regions including GovCloud IL5, ten European cities, sovereign India via GIFT City (IFSCA), thirteen APAC cities, and five Middle East cities.

Customers include a top-5 US bank, a Fortune 100 SaaS platform, a global investment bank under FedRAMP HIGH/IL7, a defense systems integrator, a large healthcare network, a multinational retailer, a managed security provider, and a Series C fintech with 1,400 microservices under management. Reported Q2 2026 transparency metrics include 1,240 active tenants, 3.42M findings shipped, 61% of findings closed by automatic patch, a 0.948 mean adversarial resistance score across the Griffin family, 147 coordinated disclosures, and 100K+ autonomous AI remediations applied. One Fortune 100 customer achieved 10,000+ autonomous fixes with mean time to remediate dropping from 11 days to 3 hours.

Name

Safeguard

Legal name

Safeguard.sh Inc

Website

https://safeguard.sh

Company type

Private

Founded year

2024

Operating status

Operating

Headcount range

51–100 employees

Short description

Safeguard is an AI-native software supply chain security platform that uses proprietary Griffin, Eagle, and Lion models to autonomously find and fix vulnerabilities across dependencies, containers, and infrastructure. It serves regulated enterprises in financial services, banking, defense, healthcare, and government with FedRAMP HIGH/IL7 sovereign deployments and 373 compliance framework mappings.

Ownership category

akta.pro rank

Product category

Application Security Software

NAICS

Computer Systems Design and Related Services (54151), Computer Systems Design Services (541512)

SIC

Services-Prepackaged Software (7372), Services-Computer Programming, Data Processing, Etc. (7370)

akta.pro primary industry

Supply Chain & Firmware Security for Industrial Devices (HDADAJAL)

akta.pro secondary industries

Security Awareness, Training & Compliance Attestation (HDADAIAJ), Configuration & Exposure Hardening (CIS/Benchmarking) (HDADAHAH)

Keywords

• Software supply chain security
• AI vulnerability remediation
• SBOM management
• Compliance automation
• Third party risk management

Headquarters

HQ city

California City

HQ country

United States

HQ region

North America

Offices4 records

Markets served

GTM type

B2B

Offering type

Software

Cost components

Technology or R&D, Personnel, Infrastructure, Operations, Marketing or Sales

Revenue model

1. Platform Subscription (Usage-Based): Usage-based subscription for the core ESSCM platform. The platform is a functional superset of point scanners (Snyk-class) and GRC tools (Vanta-class), priced at roughly a tenth of per-seat pricing. Agents can also complete the entire procurement lifecycle autonomously over MCP — evaluate, sign up, subscribe, pay — without human handoff.
2. Sovereign / Air-Gapped Deployments: Sovereign-tier and air-gapped deployments (Griffin Zero, Aegis appliance) for defense, government, and regulated enterprises. Griffin Zero runs on documented multi-GPU sizing from 11x H100 (Growth) to 22x H100 multi-AZ (Mature). Separate compute, infrastructure, and support tiers.
3. Enterprise Support & Services: Enterprise customers receive dedicated support, architecture reviews, pen-test summaries, SOC 2 reports, and full control documentation under NDA. Quarterly business reviews and private roadmap reviews are available for enterprise tiers.
4. Marketplace: Community-contributed policies, guardrails, and remediation recipes published, versioned, and consumed through the Marketplace. First wave includes 40 community-authored policies covering supply chain, AI safety, and licensing. Partner-built scanner SDK enables third-party contributions.

Pricing tiers

Go-to-market motion4 records

Distribution channels8 records

Marketing channels11 records

Core offering

Safeguard provides an AI-native software supply chain security platform (ESSCM) that autonomously discovers and fixes Zero-Day and reachability-filtered vulnerabilities across code, containers, dependencies, and infrastructure-as-code. The platform is built on a proprietary model family (Griffin for deep remediation, Eagle for adversarial triage, Lion for on-device inline analysis) and is delivered as a unified control plane that maps findings to 373 compliance frameworks and produces policy-gated pull requests.

Product overview

Safeguard is an AI-native software supply chain security platform built around autonomous remediation. The core offering is Griffin AI — a family of five purpose-built models (Griffin Lite 8B, S 14B, M 32B, L 70B, Zero 671B-MoE) — supported by Eagle (13B) for ranking and triage, and Lion (1B) for on-device inline analysis. These models are served on the Aegis architecture with sliding-window attention, a security-augmented tokeniser, and adversarial disproof heads. The platform is delivered as a unified ESSCM (Enterprise Software Supply Chain Manager) with integrated modules: Scanner Suite (11 scanners + 7 enrichment feeds), Auto-Fix for autonomous remediation, SBOM Studio for SBOM lifecycle management, TPRM for third-party risk, Open Source Manager for OSS registry intelligence, IaC Security for infrastructure scanning, and DAST for dynamic testing. Access surfaces include a web Portal, Trust Center, MCP Server (25+ tools for Claude/Cursor/Cline agents), Safeguard Code IDE agent, Safeguard Guard runtime protection, Safeguard Cowork collaboration surface, a CLI tool, and a Gold Registry marketplace of 500K+ zero-CVE hardened components. The platform covers 373 compliance frameworks with policy-gated workflows and automated evidence generation.

Differentiator

Problem solved

Functional benefit

Brands

• Griffin AI: Autonomous security agent with 100-layer reachability analysis
• Aegis Architecture
• ESSCM
• SBOM Studio
• Scanner Suite
• TPRM
• Open Source Manager
• Auto-Fix
• MCP Server
• IaC Security
• DAST
• Safeguard Code
• Safeguard Guard
• Safeguard Cowork
• Marketplace
• Gold Registry
• Griffin Lite
• Griffin S
• Griffin M
• Griffin L
• Griffin Zero
• Eagle
• Lion

Products and services

• Griffin AI Purpose-built AI reasoning engine for software supply chain security. Five variants (Lite 8B, S 14B, M 32B, L 70B, Zero 671B-MoE) perform autonomous vulnerability discovery, reachability analysis, adversarial disproof, patch authoring, and policy-gated automated remediation across the software lifecycle.
• Eagle 13B ranking and clustering model for wide-angle triage and adversarial Zero Day discovery across repositories. Eagle Gen 5 (190B) handles multi-jurisdiction policy synthesis across NIS2, DORA, EU AI Act, FedRAMP High, and ISO 27001:2022.
• Lion 1B distilled on-device model for inline gut-check in the IDE and CLI. Sub-100ms p95 on developer laptop with no network egress. Sink detection, sanitiser quality scoring, and inline triage at commit time. Works offline.
• ESSCM (Enterprise Software Supply Chain Manager) Enterprise-grade supply chain security platform providing continuous compliance monitoring across 373 frameworks. Ingests all repos, containers, and manifests into a live SBOM graph with deep transitive and cross-package taint edges, with policy-gated merge workflow from scan to fix to attestation.
• Aegis Architecture Reasoning architecture underlying every Griffin variant. Features sliding-window plus landmark attention, security-augmented tokeniser (~28k extra tokens), mixture-of-experts routing (Zero tier), and structured trace output as a first-class contract.
• Auto-Fix Autonomous remediation engine that analyses vulnerable dependencies, generates compatible upgrades, opens policy-gated pull requests, runs CI, validates regression suites, and merges fixes. Campaign mode coordinates fan-out across services affected by the same CVE.
• SBOM Studio Software Bill of Materials management for ingesting, slicing, and distributing SBOMs in CycloneDX and SPDX formats with automated generation on every build, version control for SBOM history, and EO 14028 self-attestation integration.
• TPRM (Third Party Risk Manager) Third Party Risk Manager providing vendor SBOM verification and continuous vendor tracking, with sub-processor tracking and compliance mapping across multiple regulatory frameworks.
• MCP Server Model Context Protocol server providing 25+ tenant-scoped tools enabling Claude, Cursor, Cline, and custom agents to scan repos, query vulnerabilities, generate remediation plans, apply fixes, merge PRs, and check compliance gates. Supports agentic end-to-end procurement.
• Gold Registry Marketplace of 500K+ curated zero-CVE hardened container images and packages across npm, PyPI, Maven, NuGet, Go, Rust, RubyGems, PHP, containers, and Helm, signed with SLSA L3 provenance and rebuilt continuously.
• Scanner Suite Eleven integrated scanners running in parallel against ingested artifacts, fused into a single ranked candidate set with cross-scanner evidence. Seven enrichment feeds: NVD, OSV, EPSS, KEV, GHSA, VirusTotal, VulnCheck.
• Open Source Manager (OSM) Registry intelligence for open source software dependencies. Monitors and analyses OSS packages across ecosystems, tracking maintainer risk, typosquat campaigns, and dependency confusion attacks.
• Safeguard Code Local coding agent for IDE and CLI that writes and refactors securely by default, respecting repo conventions and security policy, with 42 secure completions per hour.
• Safeguard Guard Runtime protection layer providing guardrails for LLM apps and agents, including prompt-injection, jailbreak, and exfiltration defense at the I/O boundary.
• Portal Security team's command center providing live posture monitoring, policy management, evidence trails, and compliance dashboards across 373 frameworks.
• Trust Center Public-facing live security posture page on the customer's own domain displaying real-time compliance status (SOC 2, ISO 27001, GDPR, HIPAA), controls passing rates, SBOM freshness, and verified sub-processors.
• IaC Security Infrastructure-as-Code security scanning for Terraform, Pulumi, and Crossplane, with policy gates enforced in CI and admission controller.
• DAST (Dynamic Application Security Testing) Dynamic application security testing capabilities with reviewed, ready-to-apply fix plans alongside dynamic findings.
• Safeguard Cowork Collaboration surface for security teams enabling security and engineering teams to collaborate on findings, remediation workflows, and policy management across the organisation.
• CLI Tool Command-line interface for local scanning, SBOM generation, policy enforcement, and CI/CD pipeline integration, with Lion running locally via CLI for on-device inference.

Quantifiable outcome

• ~80% reduction in false positives via reachability analysis vs traditional SCA scanners
• +7 more outcomes

Named customers10 records

Segments6 records

Ideal customer profiles6 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration27 records

AI capability11 records

Feature11 records

Partnerships

Three partnerships are on record, tiered flagship and core.

• Anthropic (Claude)flagshipTechnology or Integration · 12 June 2026Safeguard is available as a connector in Claude, bringing software supply chain security tools into Claude alongside continuous compliance monitoring. The MCP server integration enables Claude to query vulnerabilities, scan repos, generate remediation plans, apply fixes, merge PRs, and check compliance gates directly.
• All Major Frontier Model FamiliescoreTechnology or Integration · 10 June 2026Safeguard's Multi-Agent TAOR Deep Think AI Engine (now branded Griffin Gen 5) supports every major frontier model family, plus private model hosting. Universal AI model support enables continuous zero-day discovery and automated remediation wired for any model deployment.
• Marketplace Policy ContributorscoreStrategic or Co-development Partner · 12 March 2026Policies, guardrails, and remediation recipes can be published, versioned, and consumed through the Safeguard Marketplace. First wave includes 40 community-authored policies covering supply chain, AI safety, and licensing. Partner-built scanner SDK enables third-party detection logic.

Scale indicators17 records

Recent moves6 records

Expansion highlights8 records

Emerging players

• Chainguard: Closest emerging competitor on the "zero-CVE base images and dependencies" thesis. Chainguard's curated hardened container images overlap with Safeguard's Gold Registry but lack the broader SCA + GRC + autonomous remediation platform; valuable comparable for the supply-chain-clean-start story.
• Endor Labs: Emerging SCA competitor with reachability-based prioritization as its core thesis — directly comparable to Safeguard's 100-layer reachability pitch. Targets similar enterprise and FinServ buyers but lacks Safeguard's full compliance framework coverage and air-gapped sovereign tier.
• Socket: Emerging supply-chain security startup focused on malicious-package detection, dependency behavior analysis, and OSS risk in npm/PyPI. Comparable in targeting developer-led adoption with AI-driven signals; narrower in scope than Safeguard's full ESCM platform.
• Vanta: Leading automated compliance/GRC platform named explicitly by Safeguard as a displaced alternative. Overlaps on the 373-framework compliance automation and SOC 2/ISO 27001 evidence pipeline but does not provide SCA, reachability analysis, or autonomous fix authoring.

Direct peers

• Sonatype: Long-standing direct peer in software supply chain and open-source dependency intelligence (Nexus Lifecycle, Nexus Repository). Comparable focus on OSS risk and SBOM, but relies on traditional heuristics rather than a purpose-built LLM family for reachability reasoning.
• Snyk: Direct SCA/SAST competitor named explicitly by Safeguard as the incumbent it displaces. Both target developer-led adoption with per-seat pricing for code and dependency vulnerability scanning; Snyk has the larger installed base and significantly more funding, but lacks Safeguard's purpose-built AI model family and autonomous auto-fix loop.
• Veracode: Established application-security platform covering SAST, DAST, and SCA with strong regulated-enterprise penetration. Overlaps directly with Safeguard's DAST, IaC, and SCA modules and competes for the same Top-5 bank and federal-defense buyers; has a deeper historical installed base but slower model-driven roadmap.
• Mend (formerly WhiteSource): Direct SCA competitor with overlap in dependency scanning, license compliance, and container security. Targets similar regulated-enterprise and FinServ buyers with a usage-based SaaS model; differentiated from Safeguard primarily by lack of an autonomous patch-authoring agent.

Broad incumbents

• JFrog: Broad DevOps + security platform combining artifact repository (Artifactory), Xray SCA, and lifecycle management. Overlaps with Safeguard on container/IaC scanning and CI/CD gates but ships as part of a wider DevOps portfolio rather than a security-first autonomous platform.
• GitHub Advanced Security: Microsoft-owned bundled SCA/SAST/secret-scanning offering included with GitHub Enterprise. Competes head-to-head in the same developer workflow (PR-time scanning) but ships as part of a much broader platform; lacks a comparable sovereign/air-gapped tier and autonomous-remediation agent.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat6 records

Key risks5 records

Key highlights7 records

Customer concentration

• Xx.com/safeguard_sh
• GitHubgithub.com/Safeguard-sh
• LinkedInlinkedin.com/company/safeguard-sh
• Facebookfacebook.com/safeguard-sh

Compliance16 records

Revenue estimate

Valuation estimate

Number of profiles

Funding overview

Funding rounds

Investors

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Safeguard