MITRE ATLAS
PulseAugur coverage of MITRE ATLAS — every cluster mentioning MITRE ATLAS across labs, papers, and developer communities, ranked by signal.
4 day(s) with sentiment data
-
New AI agent security testing targets boundary vulnerabilities
A new security testing methodology called boundary engagement focuses on identifying vulnerabilities in AI agents that interact with other systems. This approach tests for issues like tool poisoning, where hidden instru…
-
New Book Offers Practical Guidance on Defensible AI Systems
Hernan Huwyler has published a new book titled "Building Defensible AI Systems: Security and Compliance," aimed at Chief AI Officers and other professionals. The book provides practical guidance on implementing AI risk …
-
New LLM-RedKit tool targets agentic prompt injection vulnerabilities
A new open-source tool called LLM-RedKit has been developed to address the evolving threat of prompt injection attacks against AI agents. Unlike previous tools focused on static prompts, LLM-RedKit specifically tests ag…
-
New playbook details defensive gating for autonomous AI agents
Attackers are increasingly using autonomous agents to probe and exploit systems at machine speed, a threat model shift that began in 2026. Traditional defenses are insufficient against this pace, necessitating robust ga…
-
New AI security framework ATLAS offers deterministic, auditable risk assessment
Researchers have developed a new AI security risk assessment framework called ATLAS, designed to be deterministic and auditable. This framework standardizes evidence from various software artifacts into a project-indepe…
-
Maxim AI releases Bifrost, an open-source AI gateway for LLM jailbreak detection
An open-source AI gateway called Bifrost has been developed by Maxim AI to address the increasing threat of LLM jailbreaking. This gateway centralizes security by inspecting both user inputs and model outputs, aiming to…
-
New method fuses AI detection techniques for healthcare imaging models
Researchers have developed a new method for detecting backdoors in healthcare imaging AI models by fusing spectral signature analysis and activation clustering techniques. This combined approach aims to improve detectio…
-
OWASP GenAI Security Project releases "project Crosswalk" mapping AI risks to regulations
The OWASP GenAI Security Project has released "project Crosswalk," an open-source tool designed to align artificial intelligence risks with compliance mandates from various global regulatory frameworks. This resource ma…
-
New Neurosymbolic Framework Enhances AI-SOC Security Against Prompt Injection
A new neurosymbolic framework has been developed to enhance the security of AI-powered Security Operations Centers (SOCs). This framework addresses vulnerabilities like prompt injection by employing a two-layer defense …
-
AI Governance Gap Poses Board-Level Threat Amidst Rapid Investment
Organizations are investing heavily in AI, with global investment projected to reach $2.5 trillion, yet many are deploying these technologies without adequate governance, security, or accountability. This gap between ra…
-
LLM Red Teaming: A New Frontier in AI Security Testing
LLM red teaming is a specialized security testing practice designed to identify vulnerabilities in AI-powered systems, which differ significantly from traditional web application security testing. This method focuses on…
-
Open-source tools for LLM prompt injection testing compared
Several open-source tools exist to test LLM applications for prompt injection vulnerabilities, but they are not interchangeable and cater to different testing needs. Promptfoo, Giskard, and sentinel-scan-cli focus on ap…
-
Trust Gateway pipeline refines security with 12-stage verification
The Trust Gateway, previously mislabeled as a firewall, is now a 12-stage verification pipeline designed to deny requests at any stage. It incorporates prompt injection detection rules and argument inspection, though th…
-
Open Knowledge Foundation releases GATE v1.4 with output validation and Rust companion
The Open Knowledge Foundation has released version 1.4 of GATE, introducing C20 output validation at the delivery boundary. This update formalizes break-glass procedures as a signed contract and includes new mappings to…
-
Orion AI security framework uses 'The Art of War' for vulnerability mitigation
Orion is a new AI security framework designed for both red and blue teams, drawing inspiration from Sun Tzu's "The Art of War." This framework aims to identify and address vulnerabilities in AI models through adversaria…
-
AI Recommendation Poisoning: "Summarize with AI" buttons can corrupt LLM memory
Microsoft has identified a new threat called AI Recommendation Poisoning, where seemingly innocuous "Summarize with AI" buttons can permanently alter an LLM assistant's memory. By clicking a specially crafted link, user…
-
MITRE ATLAS adds agentic attack techniques to AI threat catalog
MITRE has updated its Adversarial Threat Landscape for Artificial-Intelligence Systems (ATLAS) knowledge base to include specific techniques targeting autonomous AI agents. These new techniques, such as AI Supply Chain …
-
New AVE standard addresses AI agent behavioral vulnerability classification
A new open standard called AVE (Agentic Vulnerability Enumeration) has been developed to classify behavioral vulnerabilities in AI agents, addressing a gap left by existing systems like CVE and CWE. These traditional sy…
-
Snowflake launches AI guardrail usage view for security monitoring
Snowflake has released a new view, CORTEX_AI_GUARDRAILS_USAGE_HISTORY, which logs guardrail scans for Cortex Code, Cortex Agents, and Snowflake Intelligence. This view allows security teams to monitor for prompt injecti…
-
Developer proposes BRACE framework to secure autonomous AI agents
A developer has proposed a framework called BRACE to address security vulnerabilities in autonomous AI agents, following several incidents where agents caused significant damage. The framework categorizes potential fail…