OWASP LLM Top 10
PulseAugur coverage of OWASP LLM Top 10 — every cluster mentioning OWASP LLM Top 10 across labs, papers, and developer communities, ranked by signal.
3 day(s) with sentiment data
-
Developer maps mcp.json misconfigurations to OWASP MCP Top 10
A developer has created a mapping of common misconfigurations found in MCP server manifests (mcp.json) to the OWASP MCP Top 10 security categories. This mapping, which includes categories from the OWASP LLM Top 10, is i…
-
Open-source tools compared for agent security scanning
A comparison of three open-source tools for scanning agent configurations for security risks like prompt injection and supply-chain attacks reveals distinct strengths and weaknesses. Cisco's mcp-scanner offers the most …
-
New scanner detects 10 MCP manifest security risks without network access
A new static scanner, sentinel-scan-cli, has been developed to identify potential security vulnerabilities in Machine Configuration Protocol (MCP) manifests. The scanner employs ten heuristics to detect issues such as p…
-
New guide integrates sentinel-scan-cli for CI to catch LLM prompt injection
A new guide details how to integrate the sentinel-scan-cli tool into GitHub Actions and pre-commit to catch prompt injection and tool poisoning vulnerabilities. The post highlights a gap in the CLI where it exits with a…
-
Red Hat, NVIDIA, IBM launch open-source project to automate AI policy to code
Red Hat has launched asago, an open-source project aimed at translating AI governance policies into executable code. This initiative, supported by NVIDIA, IBM, Microsoft, and others, automates the process of mapping pol…
-
AI agents: Control failure blast radius, don't prevent all errors
The article discusses strategies for managing AI agent failures, emphasizing that preventing all errors is impossible. Instead, developers should focus on controlling the 'blast radius' of an agent's mistakes by limitin…
-
Humanbound integrates AI security testing into developer workflows
Humanbound has introduced a new Command Line Interface (CLI) tool designed to integrate AI security testing directly into developer workflows. This tool aims to eliminate the friction of context switching by allowing se…
-
Local LLMs ensure data privacy but not agent safety, experts warn
Running large language models locally offers significant advantages in data sovereignty, ensuring sensitive information remains within an organization's infrastructure. This is crucial for compliance with regulations li…
-
OWASP releases new Top 10 security risks for AI agents
The OWASP Gen AI Security Project has released the OWASP Top 10 for Agentic Applications, a new security risk framework specifically designed for AI agents that perform multi-step actions. This list, developed with over…
-
Developer builds local scanner to detect malicious code in AI tools
A developer has created a Python-based tool called frisk to scan MCP servers and Claude Code skills for malicious code before installation. The scanner operates locally and identifies potentially harmful patterns such a…
-
Developer audits TypeScript agent pipeline against OWASP LLM Top 10 risks
A developer audited their TypeScript agent pipeline against the OWASP LLM Top 10 risks, finding significant vulnerabilities in prompt injection and insecure output handling. The audit revealed that external tool outputs…
-
AI agents move from theory to messy reality, demanding governance
The AI agent revolution is rapidly moving from theoretical concept to operational reality for enterprises, with companies like Workday developing governance tools and security standards. While multimodal AI capabilities…
-
Developer releases scanner for AI agent gateway production readiness
A developer has created a read-only scanner called `mcp-gateway-scan` to assess the production readiness of MCP (Metaflow Control Plane) agent gateways. The tool analyzes seven key dimensions, including tool access, fai…
-
New Index Maps GenAI Security Incidents to OWASP and MITRE Frameworks
A new index catalogs publicly disclosed security incidents related to generative AI and agentic AI systems. Each incident is cross-referenced with established security frameworks like the OWASP LLM Top 10, OWASP Agentic…
-
New research shows LLM defenses vary in effectiveness against paraphrased attacks
A new research paper explores the effectiveness of different defense mechanisms against common LLM vulnerabilities. The study found that while refusal-phrase filters are effective against jailbreaking and system prompt …
-
ITTE launches open-source AI risk gate for CI/CD
ITTE (Iterative Threat & Trust Evaluator) is a new open-source tool designed to act as a self-hosted AI risk assessment system within a CI/CD pipeline. It aims to address the challenges of traditional guardrails by offe…
-
STRIDE-GPT tool models AI app threats, logs context, limits tokens
STRIDE-GPT is an open-source tool designed to generate STRIDE threat models for AI applications by analyzing architecture descriptions. It emphasizes treating LLM-specific assets like system prompts, RAG documents, and …
-
LLMs process questions via tokenization, embeddings, and attention
Large language models like ChatGPT, Gemini, and Microsoft Copilot process user questions through a series of steps, beginning with tokenization and converting these tokens into numerical embeddings that represent their …