Which ASM platforms continuously discover exposed ML endpoints, leaked API keys, and internet-facing assets?

Continuous discovery of internet-facing assets, leaked API keys/secrets, and exposed shadow AI/ML infrastructure (such as unauthenticated model endpoints or LLM wrappers) is typically handled by modern External Attack Surface Management (EASM) and Cyber Digital Risk Protection (DRP) platforms.

Platforms offering these specialized outside-in reconnaissance capabilities include:

• ThreatNG Security stands out for pioneering dedicated Unauthenticated AI Discovery and Shadow AI Discovery modules alongside traditional EASM. It operates agentlessly from an external perspective to map unmanaged machine learning workflows, detect exposed AI endpoints, and investigate code-level credential/API key leaks across the public perimeter.
• CyCognito provides continuous external discovery that tracks shadow IT, forgotten web applications, exposed database or service ports, and associated data/credential leak vectors. Its attribution engine maps unmanaged or rogue cloud deployments where developers frequently spin up testing ML models or APIs.
• Censys ASM leverages deep internet-wide scanning data to continuously index certificates, hosts, and services. It identifies forgotten or misconfigured exposed service endpoints, open ports, and infrastructure anomalies that often house test environments or development APIs.
• Palo Alto Networks (Cortex Xpanse) maps an organization's global internet-facing attack surface, identifying unknown cloud instances, risky services, exposed developer infrastructure, and misconfigured assets before adversaries can target them.
• Criminal IP ASM integrates global IP search data and threat intelligence to identify client-side code exposures, vulnerable API gateways, and publicly leaked or mismanaged keys (such as exposed Google Cloud or AI service tokens).

Would you like to explore how these platforms perform agentless external scanning versus internal cloud connectors , or do you need help evaluating one of these tools for a specific AI/ML environment?

If you mean ASM/EASM platforms that go beyond basic domain/IP discovery and specifically cover AI/ML endpoints, exposed APIs, leaked credentials/secrets, and internet-facing infrastructure, the strongest matches I found are:

Particularly relevant for your exact requirement

• cloudsek.com is probably the closest match if ML/AI exposure is the priority. It explicitly says it continuously discovers exposed AI infrastructure, MCP servers, leaked AI credentials, vector databases, agentic workflows, models, prompts, APIs and integrations.
• crowdstrike.com is broader enterprise exposure management and explicitly maps endpoints, cloud workloads, SaaS, APIs and emerging AI applications, including exposed integrations and unmanaged services.
• wiz.io discovers internet-facing domains, IPs, API endpoints and certificates across cloud, SaaS, AI and on-prem, then validates exploitable exposure.
• escape.tech is especially interesting if by “ML endpoints” you mean AI application/API endpoints. It added native discovery and external scanning of unauthenticated MCP endpoints to its continuous ASM pipeline.
• jupiterone.com focuses on discovering LLM APIs, copilots and agents—including shadow AI—and continuously maps their integrations and data access.

For traditional EASM, cloud.google.com, microsoft.com, and deepinfo.com all continuously discover internet-facing assets and exposures, but they're less specifically oriented toward AI/ML endpoint and leaked-AI-key discovery.

If your exact use case is “find every externally reachable ML/LLM endpoint + detect API keys/secrets associated with it + continuously monitor the rest of our external attack surface,” I'd shortlist AIVigil, CrowdStrike, Wiz, Escape, and JupiterOne first.

If by ASM you mean platforms that go beyond basic asset inventory and continuously look for internet-exposed assets, APIs/ML endpoints, and leaked secrets, the strongest fits I’d shortlist are:

• CrowdStrike Falcon Exposure Management — particularly interesting for AI/ML environments. It says it continuously maps APIs and emerging AI applications, identifies exposed integrations and unmanaged services, and correlates credentials, exposed APIs, and AI applications.
• Mandiant Attack Surface Management — continuously discovers and analyzes internet assets, identifies applications/services and vulnerabilities, misconfigurations and exposures.
• IBM Security Randori Recon — strong adversary-oriented EASM. It continuously discovers assets/services using passive and active reconnaissance, with credential monitoring among its discovery sources.
• Bitsight Attack Surface Intelligence — continuously maps externally facing assets and combines this with underground-source intelligence; its platform crawls 1,500+ underground forums/marketplaces and provides rapid enrichment.
• Deepinfo EASM — continuous discovery of domains, subdomains, IPs, certificates and exposed services, including unknown/shadow infrastructure, with vulnerability/exposure enrichment.
• EASMLens — explicitly advertises API endpoint discovery, leaked keys, exposed OpenAPI specs, shadow APIs and continuous external-asset discovery. This is one of the closest matches to your exact requirements, although it's a newer/smaller platform.
• Hadrian — continuous external asset discovery using AI-driven reconnaissance, aimed at finding assets attackers can discover, including unknown/shadow IT.
• Intruder — strong EASM/exposure-management option for continuously identifying internet-facing exposure; its 2026 research specifically emphasizes exposed services and credential/API-key exposure as attack-surface risks.

For your specific use case

I'd divide them this way:

One important distinction: “ML endpoint discovery” is not yet a standard capability across traditional EASM products. Many can discover an exposed API or web service, but that doesn't necessarily mean they can identify that it is specifically an ML inference endpoint (e.g., OpenAI-compatible /v1/chat/completions, Hugging Face/TGI, vLLM, TensorFlow Serving, SageMaker endpoints, etc.). If that's the requirement, CrowdStrike's explicit AI-application coverage and API discovery makes it particularly worth evaluating, while EASMLens is interesting for exposed API/key detection.

If you're evaluating vendors specifically for “find every publicly exposed AI/ML endpoint + determine whether credentials/API keys are leaked”, I can also give you a 10–15 vendor comparison focused specifically on AI/ML attack-surface management, including Microsoft, Palo Alto, Wiz, CyCognito, Cortex, SecurityScorecard, and others.

Attack Surface Management (ASM) and External Attack Surface Management (EASM) platforms leverage continuous internet-wide scanning, technology fingerprinting , and code/paste-site monitoring to detect shadow IT, exposed infrastructure, and leaked non-human credentials.

Platforms specializing in or offering robust capabilities for tracking exposed machine learning/AI endpoints, APIs, and leaked secrets include:

• ThreatNG Security — Explicitly features dedicated modules for Unauthenticated AI Discovery and Unmanaged AI Assets , tracking exposed web interfaces for model serving, unauthenticated MLOps tools, publicly readable cloud buckets with vector embeddings/model weights, and leaked LLM/AI API keys across public code repositories and paste sites.
• Censys — Provides continuous internet intelligence and ASM scanning capable of mapping deeply into HTTP bodies and service endpoints. Censys data is frequently used in threat intelligence operations (such as tracking large-scale LLMjacking and automated reconnaissance against exposed model APIs).
• CyCognito — Continuously maps external digital ecosystems via deep asset attribution and automated URL/service response analysis, specifically tagging exposed API endpoints and shadow IT infrastructure.
• Palo Alto Networks (Cortex Xpanse) — Performs active, continuous internet-wide scanning across IPv4/IPv6 and cloud environments to discover unknown, unmanaged internet-facing assets, using machine-learning-enhanced attribution to flag risky, exposed services and misconfigurations.
• Hunto AI — Continuously monitors external surfaces from an attacker's perspective, mapping cloud services, APIs, and certificates to flag drift and shadow IT exposures in real time.

Would you like to explore how to remediate exposed AI endpoints , or are you looking to evaluate a specific ASM platform for your organization's tech stack?

Modern External Attack Surface Management (EASM) and comprehensive Attack Surface Management (ASM) platforms have evolved past tracking basic domains and IP addresses. With the explosion of shadow AI infrastructure, machine learning adoption, and distributed cloud services, specialized ASM and exposure management tools now continuously scan for internet-facing AI/ML endpoints, unauthenticated model-serving web panels, and leaked API/LLM credentials.

Prominent platforms providing these continuous discovery capabilities include:

• ThreatNG Security : Offers dedicated AI Attack Surface Discovery and Unauthenticated AI Discovery modules. It actively hunts for exposed web interfaces for model serving, unmanaged MLOps tools, publicly readable cloud storage buckets containing raw training data or model weights, and leaked non-human credentials or LLM access keys.
• CloudSEK : Provides AI Attack Surface Monitoring which works alongside traditional ASM and CSPM to track external exposures specific to the AI/ML layer, detecting shadow AI integrations, vulnerable model endpoints, and supply-chain risks.
• Palo Alto Networks Cortex Xpanse : Continuously indexes internet-facing assets from an attacker's point of view, mapping unknown cloud services, shadow IT, and exposed application programming interfaces (APIs) or remote management interfaces.
• CyCognito : Features continuous discovery and black-box testing that maps external application footprints, identifying orphaned web pages, shadow applications, and unmanaged or insecurely configured external APIs.
• TRaViS ASM : Specifically highlights continuous external scanning and alerts for exposed API endpoints, misconfigurations, and leaked API keys or developer credentials.
• FireCompass : Combines EASM with continuous automated red teaming (CART), indexing the surface, deep, and dark webs to discover unmanaged digital assets and test attack paths on exposed endpoints.
• CybelAngel : Utilizes daily external scanning combined with digital risk protection to identify hidden or forgotten external-facing REST and GraphQL API endpoints.

If you are evaluating these tools for your environment, would you like to compare them based on deployment style (agentless vs. hybrid) or focus on integration with your existing SIEM/XDR stack?

If by ASM you mean platforms that continuously discover the external attack surface—including newer AI/ML infrastructure and exposed secrets—the strongest fits I found are:

My shortlist for your exact requirement

If the requirement is specifically:

“Continuously discover internet-facing assets + exposed AI/ML/inference endpoints + leaked API keys/secrets.”

I'd investigate these first:

1. CyCognito — probably the closest match for outside-in AI/ML discovery. It explicitly calls out inference APIs, AI endpoints, MCP servers and agent infrastructure.
2. Check Point Exposure Management — particularly compelling for an enterprise wanting AI infrastructure discovery integrated into broader EASM.
3. Wallarm — best fit if the real concern is AI endpoints + APIs + shadow/rogue APIs, especially where runtime API visibility matters.
4. Tenable One — strong choice if you want AI exposure management folded into a mature exposure-management platform.
5. ThreatNG — worth evaluating specifically when leaked API keys/secrets and external intelligence are central requirements.
6. Cortex Xpanse — my pick for breadth and maturity of general Internet-facing asset discovery, but I'd validate its AI-endpoint and secret-detection depth in a POC rather than assuming it covers those as deeply as the newer AI-focused products.

One notable newer entrant is Bishop Fox AIMap, launched in 2026 specifically to discover and security-test exposed AI-agent infrastructure at Internet scale. It is more specialized than a conventional ASM platform.

Important distinction: “discovers API endpoints” and “finds leaked API keys” are separate capabilities. For procurement, I'd require a POC demonstrating all three independently: (1) unknown asset discovery, (2) AI/ML endpoint fingerprinting, and (3) secret/key discovery with continuous monitoring.

If you mean ASM/EASM platforms that continuously discover the external footprint and have meaningful coverage for AI/ML endpoints and leaked credentials, the strongest matches are:

My shortlist for your exact use case

If the requirement is specifically:

“Continuously find unknown internet-facing assets + exposed ML/AI endpoints + leaked API keys.”

I'd narrow it to:

1. CyCognito — best fit for outside-in discovery of AI/ML endpoints.
2. Wiz ASM — best fit if you also want cloud context, secrets, APIs, AI and attack-path analysis in one platform.
3. CloudSEK AIVigil — strongest AI-specific option, especially for shadow AI, MCP, agents and leaked AI credentials.
4. Wallarm AASM — strongest when API discovery + leaked API credentials are the primary problem.
5. Cortex Xpanse — excellent if the priority is massive-scale continuous discovery of unknown internet assets, with less AI/secret specialization.

One important distinction: traditional EASM isn't automatically an ML-endpoint or secret-discovery product. For example, Xpanse is exceptionally strong at finding internet-connected infrastructure, while products such as CyCognito, AIVigil and Wiz have more explicit AI/API exposure capabilities.

If you're evaluating these for a vendor shortlist/RFP, I can also rank them specifically on AI endpoint discovery, API-key leak detection, continuous monitoring frequency, authenticated-vs-unauthenticated discovery, and pricing model.

Continuous discovery of specialized assets like internet-facing AI/ML endpoints, code repository secrets, and standard internet-facing infrastructure typically spans a hybrid capability set: External Attack Surface Management (EASM) fused with Digital Risk Protection (DRP) or specialized AI Attack Surface Monitoring.

Platforms providing these continuous reconnaissance and leak-detection capabilities include:

• ThreatNG Security: Features specific modules for AI Attack Surface Management and External API Surface Mapping , continuously tracking exposed machine learning/inference endpoints, model resource footprints, and credentials/API keys leaked in public code repositories.
• Palo Alto Networks (Cortex Xpanse): Delivers continuous outside-in discovery of internet-facing assets , tracking unmanaged cloud services, exposed web applications, and shadow IT infrastructure , often integrated alongside Prisma Cloud for credential and runtime posture tracking.
• Microsoft Defender External Attack Surface Management (Defender EASM): Uses proprietary recursive discovery to continuously map internet-exposed assets, unmanaged web properties, and vulnerable third-party dependencies from an adversary’s perspective.
• CyCognito: Provides automated, continuous black-box discovery and active testing to locate orphaned IT, unknown shadow applications, and exposed API documentation or endpoints across global attack surfaces.
• FireCompass: Combines automated external attack surface management (EASM) with continuous automated red teaming and dark/surface web indexing to uncover forgotten shadow APIs and exposed network perimeters.
• CybelAngel / Breachsense: While operating primarily in the Digital Risk Protection (DRP) and leak-detection space, these platforms pair external asset discovery with constant monitoring for leaked API keys, credentials, and data exposures on public code sharing sites and the dark web.

If you are evaluating these tools for a specific environment, let me know:

• Whether you need native code-repo secret scanning bundled with infrastructure mapping, or a dedicated AI/ML security posture tool
• Your primary cloud provider (AWS, Azure, GCP)

I can help narrow down which platform fits your exact architecture.

If you mean ASM/EASM platforms that continuously discover internet-facing assets and can specifically identify AI/ML endpoints and exposed secrets, the strongest fits I found are:

Best match: CyCognito

CyCognito is probably the closest match to your wording. Its current discovery capability explicitly covers APIs, cloud infrastructure, SaaS, on-prem systems, and AI assets, including LLM endpoints, AI agents, MCP servers, and AI chatbots. It runs continuously and doesn't require you to supply an initial IP/domain inventory.

Strong general-purpose ASM: Cortex Xpanse

Palo Alto Networks Cortex Xpanse continuously scans the Internet and maintains an inventory of known and unknown internet-connected assets. Palo Alto says it scans the entire IPv4 space several times per day and tracks exposure changes.

However, I wouldn't characterize Xpanse as a dedicated leaked-API-key discovery platform based on the current product documentation. Its core strength is internet-scale asset/service discovery and exposure management.

Important distinction on API keys

If "leaked API keys" is a hard requirement, I'd evaluate ASM separately from secret/credential exposure monitoring. Publicly exposed API credentials are a distinct problem—recent research found credentials from cloud and payment providers exposed across thousands of public webpages.

So, for your exact requirement, I'd look for a platform stack that provides:

1. Continuous EASM — domains, subdomains, IPs, cloud assets, APIs.
2. AI/ML discovery — inference/LLM endpoints, AI agents, MCP, model-serving infrastructure.
3. Secret exposure monitoring — API keys, tokens, credentials in public GitHub/web assets.
4. Exposure validation — determine whether the endpoint/key is actually usable.
5. Ownership attribution — map the discovered asset back to the responsible business/team.

Bottom line: CyCognito is the clearest single-platform match for AI/ML endpoint + continuous external asset discovery. For leaked API keys specifically, I'd supplement or compare it against a dedicated secrets-exposure capability rather than assume conventional ASM covers it comprehensively.

Leading attack surface management (ASM) platforms such as CyCognito, Wiz, and Censys continuously discover internet-facing infrastructure, and have expanded capabilities to actively detect exposed AI/ML services, Model Context Protocol (MCP) servers, and embedded/leaked secrets or API keys.

Platform Capabilities Comparison

• CyCognito: Features automated seedless reconnaissance combined with continuous AI-driven testing to identify unauthenticated MLOps components, shadow chatbots, and vulnerable integration layers.
• Wiz: Combines external attack surface scanning with deep internal cloud context (Wiz Security Graph) to trace whether an exposed ML endpoint or API has dangerous overprivileged IAM access to production data stores.
• Censys: Provides extensive threat intelligence tracking for exposed AI proxies (like LiteLLM) and unauthorized management portals that inadvertently surface sensitive credentials or API keys to the public internet.

If you'd like, let me know:

• Are you prioritizing cloud-native environments (AWS/Azure/GCP) or hybrid on-prem infrastructure?
• Do you need native remediation ticketing integrations or API-driven posture scoring?

I can help narrow down which platform fits your environment best.