Receive daily AI-curated summaries of engineering articles from top tech companies worldwide.
Endigest AI Core Summary
Databricks releases DASF v3.0, extending its AI Security Framework with 35 new risks and 6 controls specifically targeting agentic AI systems.
•New risks cover three sub-components: Agent Core (memory poisoning, goal manipulation, cascading hallucinations), MCP Server (tool poisoning, prompt injection), and MCP Client (malicious server connections, data leakage)
•The "Lethal Trifecta" concept describes peak risk when agents simultaneously access sensitive data, process untrustworthy inputs, and can change external state
•Six new controls include least privilege for tools, human-in-the-loop oversight for high-stakes actions, sandboxing of agent-generated code, AI Gateway guardrails, and agentic observability/tracing
•Multi-agent risks such as Agent Communication Poisoning and Rogue Agents are addressed for systems where agents coordinate with other agents
•The full framework now covers 97 risks and 73 controls, mapped to MITRE ATLAS, OWASP, NIST, and Cloud Security Alliance standar
This summary was automatically generated by AI based on the original article and may not be fully accurate.
Databricks
The Hacker News