Join our Newsletter — 33% off our NHI Course
Home› Glossary› Identity Beyond IAM› Bedrock Marketplace
Identity Beyond IAM

Bedrock Marketplace

← Back to Glossary
By NHI Mgmt Group Updated November 16, 2026 Domain: Identity Beyond IAM

Bedrock Marketplace is a marketplace for discovering and deploying prebuilt AI models and related assets within an AI platform. It typically lets users browse, subscribe to, and invoke models through the platform’s APIs, while platform permissions control who can access each model or capability.

What Bedrock Marketplace Is For

Bedrock Marketplace is best understood as a discovery and distribution layer inside an AI platform. It lets teams find prebuilt models and related assets, then deploy or invoke them through platform APIs, with access governed by platform permissions rather than open public availability. That makes it a convenience layer, but also a control point for who can use which model and under what conditions.

Because the marketplace sits between model discovery and runtime use, it changes how organisations think about trust. A model does not become safe or sanctioned simply because it is listed; it still needs review for provenance, permissions, usage scope, and whether its invocation path matches the organisation’s AI governance model.

For platform teams, the core question is not just what is available, but who can subscribe, who can invoke, and which identities or roles are allowed to do so. That is why marketplace design matters to both adoption speed and security boundaries.

How Marketplace Access and Invocation Work

These platforms usually separate browsing from execution. A user may be able to discover a model catalogue item, but actual invocation can depend on role assignment, project policy, API credentials, or linked service permissions. In practice, the permission model determines whether a marketplace item is merely visible, deployable, or callable in production.

That separation is important because model access is not the same as model safety. A user-facing catalogue can expose a broad set of capabilities, while the runtime controls decide whether a model can read prompts, access tools, or operate inside a larger workflow. When those boundaries are weak, discovery becomes an entry point for unintended use.

Amazon Bedrock Security Guide is the most direct companion for understanding how model access controls, service roles, guardrails, and invocation logging fit together in this platform.

Security Implications of Marketplace-Distributed Models

Marketplace-delivered models expand the attack surface because they introduce more third-party or prebuilt components into the AI stack. The main security concerns are over-permissioned access, unreviewed model behaviour, secret exposure through API usage, and the possibility that a model or related asset is invoked outside intended guardrails.

They also create a supply chain question: if an organisation consumes a model from a marketplace, it inherits some trust in the publisher, the packaging, and the surrounding metadata. If those assumptions are wrong, the risk is not only model misuse but also data leakage, cost abuse, or unauthorised downstream actions through connected APIs.

LLM Provider API Key Security and LLMjacking Guide is relevant here because marketplace access often depends on the same kinds of AI credentials and platform controls that attackers target when they want to consume models at someone else’s expense.

AI LLM hijack breach shows why model platforms need strong invocation controls: once cloud access is compromised, attackers can abuse legitimate platform access to run models or services in ways the owner did not intend.

Governance, Review, and Practical Control Points

Organisations should treat a model marketplace as part of AI governance, not just procurement. The governance burden includes deciding which models are approved, who can publish or subscribe to them, whether usage is restricted by environment, and whether logging is sufficient to reconstruct what was invoked and by whom.

The practical control points are familiar, but they need to be applied to AI assets specifically: inventory what is exposed, classify what is approved, limit who can invoke high-impact models, and monitor runtime use for anomalies. If marketplace items can be deployed quickly, then review and approval need to be equally deliberate.

NIST AI Risk Management Framework aligns well with this subject because it frames AI deployment as a risk-managed lifecycle, not a one-time catalog decision.

OWASP Non-Human Identity Top 10 also matters when marketplace usage depends on service credentials, long-lived secrets, or overprivileged machine access to models and related assets.

Operational Risks When Model Marketplaces Scale

As usage grows, the biggest risk is not a single bad model but ungoverned sprawl. Large marketplaces make it easy for teams to adopt capabilities quickly, which can lead to duplicated models, weak ownership, inconsistent permissions, and blind spots in monitoring. Over time, that weakens both security and cost control.

Another operational challenge is drift between what was approved and what is actually being invoked. A model may be subscribed for experimentation and later reused in a production workflow without fresh review. If logs, permissions, and ownership are not maintained, the organisation may lose track of which assets are active and which are merely catalog entries.

Risk and Threat Considerations

Marketplace-style AI distribution can concentrate risk when many teams rely on the same catalogue, the same permissions model, and the same credential paths. If an attacker compromises those access paths, they may gain broad model invocation capability, secret exposure, or the ability to abuse platform trust at scale.

Failure mechanism: weak approval controls, overbroad permissions, or exposed API credentials can let an unauthorised user invoke models, reuse approved access in unintended contexts, or pivot from discovery into runtime abuse.

Impact: the result can be unauthorised model use, data leakage through prompts or outputs, unexpected spend, loss of governance over deployed AI assets, and in some cases broader platform compromise if the same access path reaches adjacent services.

Standards & Framework Alignment

This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.

NIST SP 800-53 Rev 5 provides the primary governance reference for this term.

FrameworkControl / ReferenceRelevance
NIST SP 800-53 Rev 5IA-5 — Authenticator ManagementMarketplace use depends on managed credentials and invocation access.
AC-6 — Least PrivilegeModel browsing, deployment, and invocation require different access boundaries.
AU-2 — Event LoggingModel invocation and subscription actions need traceable records for review.
Recommendation — Rotate and control authenticators used to access AI platform APIs and model services. Restrict marketplace and model invocation rights to the minimum necessary roles. Log marketplace access, model invocation, and administrative actions for auditability.

Practitioner Guidance

Governance implication: treat marketplace entries as controlled AI assets, not just software listings. The permission model should distinguish browse, subscribe, deploy, and invoke rights, because those actions carry different risk and ownership requirements.

What to watch for: look for reused credentials, broad service roles, and assets that are visible in the marketplace but poorly tracked after deployment. If the organisation cannot answer who can invoke a model and from where, the control model is too loose for production use.

Practitioner takeaway: a secure AI marketplace depends less on the catalogue itself than on the identity, permission, and logging layer that governs what happens after discovery.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Reviewed and updated by the NHIMG editorial team on November 16, 2026.
NHI Mgmt Group — the #1 independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org