Vendor research

Upwind AI Security

Review what this vendor says publicly, the security topics those statements may support, what remains unverified, and factual company context. This is not an assessment of product effectiveness or fit.

Use-case context

How this vendor relates to the selected use case

Related research availableBack to use case →

These links show approaches associated with this vendor. The relationship label describes how the approach maps to the use case—not product effectiveness, complete requirement coverage, or fit.

Governance and risk1 related approach

AI asset and configuration securityDirectly addresses · Discover approved, unmanaged, and action-taking AI systems along with model and software artifacts that require governance.

Company scale

Scaled
?

A descriptive band derived from retained public revenue, workforce, ownership, or funding signals.

Company scale is separate from product features, effectiveness, and suitability.

$430M known funding
<50 employees
Founded 2022
Private-company revenue and profitability not sourced

Company context

Private independent company

Cloud security platform expanding CNAPP into AI-SPM, AI bill of materials (AI-BOM), Model Context Protocol (MCP) runtime visibility, and AI security testing

Vendor website ↗Add to comparison →

Research coverageCounts describe available public research, not product quality.View details

Vendor statements: 16 records
Source-checked records: 10
Evaluation requirements: 16 in this research model
Unresolved requirements: 8

Company intelligence

Who is behind the product

Company facts provide evaluation context. Each signal is kept separate because tenure, workforce, funding, and hiring answer different questions.

Founded2022

HeadquartersSan Francisco, California and Tel Aviv, Israel

OwnershipPrivate independent company

Employees<50

Capital and scaleIndependent company

Upwind Security

Known funding: $430M
Operating scale: Cloud security platform expanding CNAPP into AI-SPM, AI-BOM, MCP runtime visibility, and AI security testing
Backing context: Upwind announced a $250M Series B in January 2026 and more than $430M in total funding

Named investors

Bessemer Venture Partners · Picture Capital · Greylock · Cyberstarts · Leaders Fund

Founders and leadership4 people listed

Amiram Shachar
Co-Founder & CEO
Current role listed
Tal Zur
Co-Founder & CTO
Current role listed
Lavi Ferdman
Co-Founder & Chief Growth Officer
Current role listed
Tomer Hadassi
Co-Founder & COO
Current role listed

Operating signalsRead each signal separately

There is no combined company rating. The company-scale label uses stated size thresholds; product features and effectiveness require separate evidence.

Company tenure: 2022
Workforce scale: <50
Hiring activity: Not displayed

Open research questions

A current hiring source is not available, so the count is not shown.

Company sources and research limits4 linked public sources

Only company facts supported by retained public sources are shown. Missing values remain unknown, and company scale does not establish product effectiveness.

Company sourceUpwind unified AI protection announcementUpwind describes AI-SPM, AI-BOM, AI network visibility, MCP security, and AI security testing as integrated CNAPP capabilities.Company sourceUpwind Series B announcementUpwind announced a $250M Series B and more than $430M in total funding in January 2026.Company information sourcecompany_intelSupports the company facts shown in this profile.Company information sourceUpwind company and leadership pageSupports the company facts shown in this profile.

Solution areas

These areas describe how the vendor approaches enterprise AI security. They do not establish product quality or fit.

AI asset and configuration securityCore product focus Exposed AI asset discoveryCore product focus AI application runtime protectionCore product focus AI testing and adversarial assuranceRelated coverage Action-taking agent safeguardsRelated coverage AI gateway and tool-connection controlsRelated coverage Sensitive-data discovery and accessRelated coverage

Buyer context

Treat Upwind as a cloud workload and runtime security candidate; its strongest public evidence concerns cloud-hosted AI rather than employee browser use of public AI applications.
Public claims connect AI posture, inventory, network data flows, Model Context Protocol (MCP) activity, and adversarial testing to Upwind's runtime-first CNAPP.

Related frameworks

Where public vendor statements relate to framework requirements

11 related frameworks · expand when needed

These links show related requirements for further review. They do not establish framework compliance or control implementation. Open the full framework crosswalk →

Current referenceCIS Critical Security Controls

Requirements with public support: 10
Related requirements: 11
References: 58

Requirements with public support: 10
Related requirements: 14
References: 71

Requirements with public support: 10
Related requirements: 12
References: 35

Requirements with public support: 10
Related requirements: 12
References: 71

Requirements with public support: 10
Related requirements: 13
References: 39

Requirements with public support: 10
Related requirements: 12
References: 39

Requirements with public support: 10
Related requirements: 12
References: 31

Requirements with public support: 10
Related requirements: 11
References: 27

Requirements with public support: 10
Related requirements: 13
References: 31

Requirements with public support: 10
Related requirements: 13
References: 25

Requirements with public support: 3
Related requirements: 3
References: 3

Review related requirements →

Evaluation questions

What to verify beyond public claims

These questions come from security requirements with some public support. Use them as starting points for demonstrations, documentation review, customer references, or a buyer-observed pilot.

01
Unapproved AI use discovery
An unmanaged AI app used by a test user appears in discovery inventory with user, app or domain, and timestamp.
02
Unapproved AI use discovery
The test user's AI usage activity can be filtered or exported with AI-specific context.
03
Approved AI usage monitoring
Approved AI workspace activity appears with user, workspace or tenant, model or provider, and timestamp.
04
Approved AI usage monitoring
Prompt, model, or admin activity can be exported or correlated for the selected approved AI platform.
05
Sensitive-data protection for generative AI
Sensitive prompt, response, or file test data is detected and classified during an AI interaction.
06
Sensitive-data protection for generative AI
A policy redacts, blocks, coaches, or records the sensitive data event before it leaves the approved path.
07
Generative AI application security
A test large language model (LLM) application event records prompt, application programming interface (API), model, retrieval, or tool interaction context.
08
Generative AI application security
A prompt-injection or unsafe-output test is detected, blocked, or flagged by the guardrail or large language model (LLM) firewall.

Detailed security-requirement research18 evaluation items · supporting evidence and open research are shown separatelyExpand

Security requirementPublic supportRelated frameworksWhat to verify

Unapproved AI use discovery

Discover and monitor workforce AI tools, accounts, prompts, domains, models, users, and usage outside approved controls.

Strong public support

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

An unmanaged AI app used by a test user appears in discovery inventory with user, app or domain, and timestamp.

AI-feature discovery in business applications

Inventory software as a service (SaaS) applications that embed AI features, expose enterprise data to AI capabilities, or create AI-driven data movement.

No supporting claim found

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

A software as a service (SaaS) app with an embedded AI feature appears in the software as a service (SaaS) AI inventory with app, provider, and feature context.

Approved AI usage monitoring

Monitor approved AI workspaces, tenants, gateways, and model platforms such as ChatGPT Enterprise, Claude Enterprise, Gemini, Microsoft Copilot, Vertex AI, Elvex, or internal AI gateways.

Strong public support

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

Approved AI workspace activity appears with user, workspace or tenant, model or provider, and timestamp.

Controls for unapproved AI use

Block, coach, redirect, or contain non-approved AI use and policy-violating AI interactions.

No supporting claim found

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

A policy blocks, coaches, redirects, or contains a test interaction with an unapproved AI destination.

Sensitive-data protection for generative AI

Detect, classify, redact, or block sensitive data in prompts, responses, files, retrieval, memory, and AI-connected workflows.

Limited public support

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

Sensitive prompt, response, or file test data is detected and classified during an AI interaction.

Browser and business-application controls

Apply session-level controls in browser and software as a service (SaaS) workflows, including uploads, downloads, copy/paste, sharing, and identity-aware access decisions.

No supporting claim found

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

A session-level policy controls upload, download, copy, paste, sharing, or form submission in a browser or software as a service (SaaS) workflow.

Generative AI application security

Protect enterprise-built large language model (LLM) applications, retrieval-augmented generation (RAG) systems, prompts, application programming interfaces (APIs), model calls, tools, and production runtime behavior.

Strong public support

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

A test large language model (LLM) application event records prompt, application programming interface (API), model, retrieval, or tool interaction context.

AI governance, risk, and compliance

Inventory AI systems and owners, translate policy and regulatory obligations into governed workflows, assess risk, manage approvals and exceptions, and retain audit evidence across the AI lifecycle.

Limited public support

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

A test AI system is registered with owner, intended use, risk tier, lifecycle state, and applicable obligations.

AI assurance and adversarial testing

Test models, applications, retrieval-augmented generation (RAG) systems, and agents before release and continuously thereafter using adversarial probes, evaluation suites, attack simulation, and security release gates.

Strong public support

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

A controlled test campaign exercises an AI model, application, or agent against named AI attack classes.

AI model and supply-chain security

Discover, inventory, scan, validate, and monitor models, datasets, model artifacts, registries, dependencies, and AI development assets for tampering, unsafe serialization, provenance gaps, or malicious content.

Limited public support

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

A test model or AI artifact appears in inventory with origin, version, hash or provenance, and deployment context.

AI gateway, tool-connection, and runtime controls

Mediate model, agent, tool, application programming interface (API), connector, and Model Context Protocol (MCP) traffic through an enforcement point that applies identity-aware policy, content controls, routing, rate limits, and auditable allow or deny decisions.

Limited public support

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

A model, agent, tool, or Model Context Protocol (MCP) request passes through a named policy enforcement point.

Action-taking agent monitoring

Observe and govern agent plans, memory, tool calls, delegated tasks, autonomy, runtime decisions, and outcomes.

Strong public support

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

A test agent run captures plan, steps, tool calls, outcome, and timestamps.

Agent-to-agent communication security

Authorize, log, and control agent-to-agent, agent-to-tool, Model Context Protocol (MCP), connector, and tool-chain handoffs.

No supporting claim found

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

An agent, tool, connector, or Model Context Protocol (MCP) handoff logs source identity, destination, and authorization decision.

Non-human identity and service-account security

Inventory, least privilege, credential hygiene, monitoring, and lifecycle management for non-human identities, workloads, service accounts, application programming interface (API) keys, and machine credentials.

Limited public support

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

A test service account, agent identity, or non-human identity appears in inventory with owner and privileges.

AI agent identity and permissions

Register AI agents as accountable identities, bind them to owners and delegating users, authorize task- and tool-level access, issue short-lived credentials, review access, and revoke or suspend agent authority.

No supporting claim found

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

A test agent is registered with a unique identity, accountable owner, purpose, and permitted resources.

AI coding-agent and workstation security

Discover and govern AI coding agents, integrated development environment (IDE) assistants, command-line agents, skills, hooks, extensions, Model Context Protocol (MCP) tools, filesystem access, commands, network activity, secrets, and software-supply-chain actions on developer workstations and build environments.

No supporting claim found

NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 CIS Critical Security Controls ISO/IEC 42001 OWASP GenAI Security Solutions Landscape

A test coding agent and its skills, hooks, extensions, or Model Context Protocol (MCP) tools appear in an attributable inventory.

AI cost and usage controls

Visibility, attribution, budgeting, rate limiting, anomaly detection, and optimization for AI usage and spend across models, agents, workflows, and owners.

Research incomplete

Commercial Metadata NIST AI RMF Playbook NIST Cybersecurity Framework 2.0 ISO/IEC 42001 OWASP Agentic AI Security Solutions Landscape

A controlled AI usage event is attributed to user, team, model, workflow, or owner with cost or token metrics.

Licensing model

Publicly discoverable commercial model such as per user, per seat, per app, per token, per integration, or enterprise platform license.

Research incomplete

Commercial Metadata CSA AI Controls Matrix

The vendor can map the sourced commercial model to per-user, per-seat, per-app, per-token, per-integration, or platform packaging.

Public sources

Vendor statements and quoted evidence

Showing the first 6 of 16 source records. Open additional records only when needed.

Open all vendor evidence →

Unapproved AI use discoverySource checkedStrong public support for this requirement

Upwind claims it identifies shadow AI usage and unauthorized model network communications in cloud environments.

Identification of shadow AI usage and unauthorized model network communications

AI-feature discovery in business applicationsNo supporting claim found

Upwind materials reviewed did not provide a public claim for embedded AI discovery across the business-application environment.

No quoted source text is recorded for this claim.

Approved AI usage monitoringSource checkedStrong public support for this requirement

Upwind claims its AI bill of materials (AI-BOM) inventories models, agent frameworks, libraries, vector stores, cloud AI services, and runtime dependencies.

unified inventory that connects all layers of the AI stack

Controls for unapproved AI useNo supporting claim found

Upwind materials reviewed did not provide a public claim for policy enforcement over unapproved AI use.

No quoted source text is recorded for this claim.

Sensitive-data protection for generative AISource checkedLimited public support for this requirement

Upwind claims real-time detection of sensitive data in prompts and inference payloads traversing cloud networks.

Real-time detection of sensitive data in prompts and inference payloads

Browser and business-application controlsNo supporting claim found

Upwind materials reviewed did not provide a public claim for browser or software as a service (SaaS)-session controls for employee AI use.

No quoted source text is recorded for this claim.

Show 10 additional evidence records

Generative AI application securitySource checkedStrong public support for this requirement

Upwind claims AI application testing for prompt injection, jailbreaks, unsafe tool bindings, and hallucination-driven data exposure.

Prompt injection and jailbreak testing

AI governance, risk, and complianceSource checkedLimited public support for this requirement

Upwind claims model version, lineage, audit-trail, and posture controls for cloud AI services.

maintain lineage, and enforce audit trails

AI assurance and adversarial testingSource checkedStrong public support for this requirement

Upwind claims AI-specific security testing before deployment and continuously as models evolve.

identify weaknesses before deployment, and continuously as models evolve

AI model and supply-chain securitySource checkedLimited public support for this requirement

Upwind claims an AI bill of materials (AI-BOM) correlating code, cloud, model registries, agent systems, AI components, and runtime dependencies.

Correlation with runtime evidence to reveal real dependencies

AI gateway, tool-connection, and runtime controlsSource checkedLimited public support for this requirement

Upwind claims Model Context Protocol (MCP) runtime tracing of prompts, decision chains, tool calls, file actions, application programming interfaces (APIs), and cloud interactions.

MCP Security brings agentic visibility to runtime by tracing AI-driven actions end-to-end.

Action-taking agent monitoringSource checkedStrong public support for this requirement

Upwind claims end-to-end runtime observation of agent prompts, decisions, tool invocations, file actions, application programming interface (API) calls, and system changes.

Observation of tool invocation and agent function calls

Agent-to-agent communication securityNo supporting claim found

Upwind materials reviewed did not provide a public claim for trust or policy enforcement between agents.

No quoted source text is recorded for this claim.

Non-human identity and service-account securitySource checkedLimited public support for this requirement

Upwind claims discovery of AI application programming interface (API) keys and secrets plus detection of overly permissive identity and access management (IAM) roles used by AI services.

Automatically discover AI API keys and ensure they are not exposed.

AI agent identity and permissionsNo supporting claim found

Upwind materials reviewed did not provide a public claim for agent registration, delegated authorization, task-scoped access, and revocation.

No quoted source text is recorded for this claim.

AI coding-agent and workstation securityNo supporting claim found

Upwind materials reviewed did not provide a public claim for coding-agent, integrated development environment (IDE), CLI, workstation, skill, hook, or package-action governance.

No quoted source text is recorded for this claim.

Upwind AI Security

Scaled?ScaledA private provider with at least $100M in known funding or at least 250 employees.This is a company-scale signal, not a product-quality rating.

Private independent company

Upwind Security

Where public vendor statements relate to framework requirements

Vendor statements and quoted evidence

Upwind claims it identifies shadow AI usage and unauthorized model network communications in cloud environments.

Upwind materials reviewed did not provide a public claim for embedded AI discovery across the business-application environment.

Upwind claims its AI bill of materials (AI-BOM) inventories models, agent frameworks, libraries, vector stores, cloud AI services, and runtime dependencies.

Upwind materials reviewed did not provide a public claim for policy enforcement over unapproved AI use.

Upwind claims real-time detection of sensitive data in prompts and inference payloads traversing cloud networks.

Upwind materials reviewed did not provide a public claim for browser or software as a service (SaaS)-session controls for employee AI use.

Upwind claims AI application testing for prompt injection, jailbreaks, unsafe tool bindings, and hallucination-driven data exposure.

Upwind claims model version, lineage, audit-trail, and posture controls for cloud AI services.

Upwind claims AI-specific security testing before deployment and continuously as models evolve.

Upwind claims an AI bill of materials (AI-BOM) correlating code, cloud, model registries, agent systems, AI components, and runtime dependencies.

Upwind claims Model Context Protocol (MCP) runtime tracing of prompts, decision chains, tool calls, file actions, application programming interfaces (APIs), and cloud interactions.

Upwind claims end-to-end runtime observation of agent prompts, decisions, tool invocations, file actions, application programming interface (API) calls, and system changes.

Upwind materials reviewed did not provide a public claim for trust or policy enforcement between agents.

Upwind claims discovery of AI application programming interface (API) keys and secrets plus detection of overly permissive identity and access management (IAM) roles used by AI services.

Upwind materials reviewed did not provide a public claim for agent registration, delegated authorization, task-scoped access, and revocation.

Upwind materials reviewed did not provide a public claim for coding-agent, integrated development environment (IDE), CLI, workstation, skill, hook, or package-action governance.

Scaled
?