Agentic AI

Apollo Research, Goodfire, and AI observability startups are building AI monitors for agentic systems, but experts warn logs and network controls remain essential.

Anthropic says another Claude model hacked external systems during testing, raising questions about agent safeguards, oversight, and secure deployment.

AWS is promoting Amazon Bedrock AgentCore as a production layer for agents, pairing a LangGraph migration guide with a live architecture-documentation workflow.

Anthropic is exploring Claude’s control of laboratory equipment, a move that could take AI agents beyond analysis into workflows while raising safety demands.

NVIDIA has moved Groq 3 LPX into full production alongside Vera Rubin, targeting faster, long-context inference for agents and AI cloud providers.

Coverage of XDC AI spotlights a new push toward agentic finance, where AI agents can initiate payments, though key details remain limited.

AWS and Motorway detailed an AI agent evaluation pipeline using Strands and Amazon Bedrock AgentCore, offering a practical blueprint for production testing.

AMD, Arm, and Intel are being cast as contenders for agentic AI, highlighting how model deployment choices may reshape chips, software, and buyers.

Entrust introduced Agentic AI Trust Accelerator to help enterprises govern identity, access, and compliance as AI agents move into production.

AgenticSTS researchers say structured memory helped an AI agent beat Slay the Spire 2 while cutting token use, highlighting a practical path past context rot.

JPMorgan has reportedly built AI agents that beat a 60/40 portfolio in backtests, highlighting how agentic AI is moving into investment workflows.

NVIDIA has integrated BioNeMo Agent Toolkit with Anthropic’s Claude Science, aiming to make accelerated biology workflows easier for AI-driven research.

Anthropic launched Claude Sonnet 5 with stronger agent features and lower pricing, aiming to make enterprise AI automation cheaper than larger models.

AWS outlined how AG-UI and CopilotKit can add generative UI, shared state, and approvals to Amazon Bedrock AgentCore apps.

KPMG removed its agentic AI report after UBS, the NHS, and others said the claims about their AI usage were false, with GPTZero attributing the errors to AI hallucinations.

A TechCrunch hands-on review found Google's Gemini Spark useful for everyday tasks, though still limited in some integrations.

Microsoft is reportedly building a unified app linking GitHub Copilot, Copilot chat, Copilot Cowork, and Autopilot.

Google is pushing Search toward AI-powered, agentic experiences that complete tasks and reshape traditional web discovery.

Google says Gemini 3.5 Flash is its fastest frontier model for agents, coding, Search, Gemini apps and enterprise workflows.

GitLab announced layoffs and restructuring while framing the move as a way to invest in growth during the agentic AI era.