Decrypt

This Half-Gigabyte AI Model Runs Local Agents on Your Phone

OpenBMB's 1B-parameter model MiniCMP 5 brings MCP support and agentic tool use to on-device AI—but it has trouble with logic ...
IEEE

Security and Quality in LLM-Generated Code: a Multi-Language, Multi-Model Analysis

Abstract: Artificial Intelligence (AI) driven code generation tools are increasingly used throughout the software development lifecycle to accelerate coding tasks. However, the security of ...
Tech Times

llama.cpp GGUF Parser Flaws: Critical Integer Overflow Enables Arbitrary Reads in Every Local AI Stack

GGUF parser vulnerabilities disclosed May 15, 2026 include a critical integer overflow that lets any malicious model file trigger arbitrary memory reads — affecting Ollama, LM Studio, and every local ...
MUO on MSN

Local LLM setup: how to use RAG and an embedding model to stop wasting context

Local LLMs degrade fast when context fills up. An embedding model and RAG pipeline fixes that — and runs entirely on your ...

A 0.12% parameter add-on gives AI agents the working memory RAG can't

Researchers built delta-mem to give AI agents working memory at 0.12% parameter overhead, outperforming RAG and context ...

Forcepoint details TeamPCP supply chain attack that turned LiteLLM into a credential stealer

A new report out today from cybersecurity company Forcepoint LLC’s X-Labs research team details a supply chain attack that ...
TestingCatalog

OpenSquilla launches open-source AI agent to cut token costs

OpenSquilla is an open-source Python AI agent with ML model routing, four-tier memory, and syscall-level sandbox isolation.
Las Vegas Sun

CoreWeave Sandboxes Launches to Accelerate Reinforcement Learning, Agent Tool Use, and Model Evaluation

The Essential Cloud for AI™, today announced CoreWeave Sandboxes, an execution layer that gives AI researchers and platform teams secure, isolated environments for running reinforcement learning (RL), ...
Martin Cid Magazine

An AI wrote a working zero-day exploit — Google caught it first

Google's Threat Intelligence Group says a criminal hacker group used a large language model to find a previously unknown flaw in a widely used open-source admin ...
eWeek

This AI Agent Is Quietly Leaving OpenClaw in the Dust

Hermes Agent’s latest release shows how AI agents are evolving from assistants into self-improving tools that learn, build, ...

MongoDB announces platform enhancements for enterprise-ready AI production

The company announced the availability of MongoDB 8.3, building on previous generations of the database software with superior performance aimed at the agentic AI era. To support this, MongoDB added ...