What happened
JFrog announced advancements to its AI Security platform, introducing Local Shadow AI Detection to inventory AI assets on developer endpoints and AI Asset Scanning with Bring Your Own LLM Key (BYOK) to vet skills and plugins using customer-controlled models.
JFrog, a provider of software supply chain platforms, announced new features for its AI Security suite during the swampUP Europe 2026 event. The primary addition is Local Shadow AI Detection, which extends the company's existing detection capabilities to developer endpoints. This feature inventories AI assets, such as agents, skills, and plugins, running directly on developer machines, areas where traditional pipeline and registry controls do not reach.
The company also introduced AI Asset Scanning with Bring Your Own LLM Key (BYOK). This capability allows customers to scan AI skills, coding agent plugins, and APM packages using their own private LLM instances. The initial release supports Anthropic’s Opus 4.8, OpenAI’s GPT 5.6 Sol, and Google’s Gemini 3.1 Pro. This approach enables organizations to vet the AI supply chain on infrastructure they control, addressing data privacy and compliance concerns.
JFrog cited its 2026 Software Supply Chain Security State of the Union report, which identified 969 malicious AI agent skills carrying high-impact payloads and 495 malicious AI models on Hugging Face. The company argues that existing application security tools are designed to catch vulnerable code, whereas AI assets often consist of plain-English instructions that agents execute, requiring a different scanning approach focused on intent.
Why it matters
The release addresses a critical gap in enterprise security where AI agents and plugins operate outside traditional pipeline controls. By extending visibility to developer machines and allowing semantic scanning via private LLMs, JFrog provides a mechanism for organizations to enforce zero-trust policies on agentic software components, which are increasingly used in development workflows.
The announcement highlights a growing security risk: developers are installing AI agents and plugins through various channels like Git, Slack, and marketplaces, often without security team review. These assets run with the developer's permissions and credentials, creating a blind spot for traditional security operations.
By integrating detection, scanning, and policy enforcement into a single system of record, JFrog aims to help security teams prove what AI agents have touched in the supply chain. This is particularly relevant for regulated industries that require strict compliance and audit trails for software components.
The BYOK feature is significant because it allows enterprises to use their own LLM infrastructure for scanning, potentially reducing reliance on third-party cloud services for sensitive security analysis. This aligns with broader trends in where organizations seek to maintain control over their AI tooling.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').An agent must create a draft calendar event for Tuesday at 2 p.m. Which evidence would establish the requested result?
What to watch next
Monitor adoption rates among Fortune 100 clients and whether other security vendors adopt similar endpoint-level AI asset detection. Watch for updates on the specific LLM models supported in the BYOK feature and any reported false positives or performance impacts on developer workflows.
The availability of these features is currently limited to JFrog Unified Security and Ultimate Security customers. The impact will depend on how widely these enterprise tiers are adopted and whether the features are eventually made available to broader customer segments.
The effectiveness of the Local Shadow AI Detection will be tested by its ability to accurately identify and inventory the diverse range of AI assets developers use, including MCP servers and hooks, without causing significant friction in the development workflow.
Competitors in the AI security space may respond with similar endpoint detection capabilities or enhanced scanning features. The market for AI agent security is evolving rapidly, and JFrog's move positions it as a key player in the 'AgentSecOps' domain.