xAI & Grok

Of the original 11 co-founders who started xAI with Musk, only two remain; xAI co-founders Zihang Dai and Guodong Zhang left after Musk complained the company's AI coding tools were not competing with Claude Code or Codex.

"Macrohard," xAI's effort to build an AI agent capable of doing anything a white collar worker can do, has been put on "pause" as its leader Toby Pohlen left weeks after being appointed.

Grok Voice Agents perform tasks and lookup information in real time; via API, developers can integrate custom tools or access xAI's real-time search, with multiple expressive voices including Ara, Eve, and Leo.

The DoD signed an agreement with xAI to use Grok in classified networks; this followed the Pentagon's decision to label Anthropic a supply chain risk.

Andrew Milich and Jason Ginsberg from AI coding tool company Cursor are joining xAI, where the two held joint responsibility for product engineering.

Musk said the purpose of the SpaceX acquisition is building "orbital data centers," which are the most cost-effective way of producing AI computing power.

Grok Voice is the most intelligent voice agent driving real-time conversations and available via agent API; it delivers low-latency speech in dozens of languages with tool calling and real-time data access, serving millions of users across the Grok mobile app and Tesla vehicles.

OpenAI Agents

OpenAI introduced Codex Security, an AI agent designed to identify cybersecurity vulnerabilities and propose fixes. The tool launched in research preview on March 6, 2026, for ChatGPT Enterprise, Business and Edu customers, with free usage for one month. The agent combines agentic reasoning with automated validation to deliver high-confidence security findings.

OpenAI unveiled GPT-5.4 Thinking, a frontier model that integrates reasoning, coding, and agentic workflows. The model features a 1-million-token context window, autonomously executes multi-step workflows, and scored 75% on OSWorld-V benchmark for desktop productivity tasks, above the human baseline of 72.4%.

OpenAI released Codex app for Windows, enabling parallel agent execution with isolated worktrees. The native Windows application supports Skills, Automations, and Worktrees, using PowerShell sandbox with no WSL requirement, available via Microsoft Store.

OpenAI secured a record $110 billion funding milestone led by Amazon ($50B), SoftBank ($30B), and Nvidia ($30B). The funding includes a $100 billion AWS collaboration for AI infrastructure, supporting enterprise agent deployment and scaling capabilities.

Google DeepMind's AlphaEvolve, a Gemini-powered coding agent, has been deployed inside Google infrastructure for over a year, recovering 0.7% of worldwide computing resources and speeding up a key Gemini kernel by 23%. The agent discovered new mathematical structures improving state-of-the-art complexity theory results.

OpenAI released interactive visual learning modules in ChatGPT covering 70+ math and science topics, allowing real-time formula and variable experimentation. Feature supports hands-on exploration of concepts like Pythagorean theorem and ideal gas law.

Anthropic deployed Claude Sonnet 4.6 and Opus 4.6 with 1-million-token context windows in early March 2026. Opus 4.6 was simultaneously deployed as add-ins in Microsoft PowerPoint and Excel, integrating AI agents into enterprise productivity workflows.

Anthropic & Claude

Anthropic announces Claude Partner Network program with $100 million 2026 commitment for enterprise adoption support, including training, certification programs, and dedicated technical support. Partners include Accenture (training 30,000 professionals) and Deloitte.

Anthropic launches Claude Opus 4.6 (most intelligent model for complex agentic tasks) and Sonnet 4.6 (balanced model with improved agentic search) both supporting 1M token context window in beta and adaptive thinking.

Claude Code receives major updates including voice mode (push-to-talk), /loop command for recurring tasks, 1M context window, Agent Teams feature for multi-agent repository work, and Opus 4.6 as default model.

New ability to create and schedule both recurring and on-demand tasks in Cowork, plus new Customize section grouping skills, plugins, and connectors for agent workflow automation.

Enterprise Analytics API provides programmatic access to usage and engagement data for Claude and Claude Code Remote usage within organizations.

Starting from Claude Code v2.1.76, MCP servers can request structured input during task execution, enabling interactive forms and browser-based data collection without workflow interruption.

Anthropic launches Code Modernization starter kit for Claude partners, providing straightforward starting point for migrating legacy codebases and remediating technical debt using Claude's agentic coding capabilities.

Microsoft Copilot

Microsoft announced Wave 3 of Microsoft 365 Copilot featuring Copilot Cowork, combining Anthropic's Claude model for multi-step task automation with enterprise data protection. Research preview available through Frontier program in March.

Microsoft launched Copilot Cowork built with Anthropic to enable AI agents completing entire tasks autonomously. Claude model now available across full Copilot Chat experience. Shift from OpenAI-only approach to multi-model flexibility.

Beginning April 1, 2026, Microsoft expanding Copilot + Power Accelerator to include Agent 365 across immersion briefings and deployment accelerators. Agent 365 added as eligible workload for CSP incentives.

Copilot Studio SaaS platform enabling autonomous agents executing tasks independently. Features Copilot Tuning for fine-tuning models and advanced Microsoft 365 Copilot integration released during 2025 Wave 2 (Oct 2025 - March 2026).

Microsoft's Frontier Transformation initiative positions AI agents as embedded capabilities in Microsoft 365 workflow. Emphasizes governance, transparency, and enterprise-grade controls with Work IQ context layer.

Google & Gemini

Google Cloud released its 2026 AI Agent Trends Report forecasting multi-agent collaboration, automated complex workflows, concierge-style customer service, and enhanced security operations through agentic AI.

Google introduced the Universal Commerce Protocol (UCP) for agent-driven shopping experiences, enabling personalized checkout directly on Google Search and Gemini, compatible with Agent2Agent and Model Context Protocol standards.

Google published a guide on eight essential multi-agent system design patterns with sample code for Google's Agent Development Kit, covering sequential pipelines to human-in-the-loop architectures.

Agent Frameworks

LangChain announced a comprehensive integration with NVIDIA combining LangSmith, LangGraph, Deep Agents, and NVIDIA Agent Toolkit to deliver an enterprise-grade agentic AI development platform with production-ready features including NVIDIA OpenShell for secure agent sandboxing, NIM microservices, and AI-Q Blueprint.

Deep Agents framework shipped with MIT licensing, scoring 42.65% on Terminal Bench 2.0 using Claude Sonnet 4.5. The open-source agent harness is available via pip install deepagents and includes planning, subagents, persistent memory, and virtual filesystem capabilities built on LangGraph.

Technical deep dive into LangGraph as a stateful orchestration framework used by production companies like Klarna, Replit, and Elastic. Demonstrates multi-agent system architecture with state persistence, human approval gates, and complex workflow orchestration across specialized agents.

Official announcement of LangChain-NVIDIA integration featuring NVIDIA-optimized execution strategies including parallel execution and speculative execution for LangGraph workflows, with NIM deployment achieving up to 2.6x higher throughput and LangSmith processing over 15 billion traces.

LangChain's 2026 State of Agent Engineering report shows 57.3% of organizations now have agents in production, up from 51% prior year. Customer service (26.5%) and research/data analysis (24.4%) are dominant use cases, with custom agents built on LangChain and LangGraph being widely adopted.

Agent Server platform received multiple March 2026 updates including enhanced security measures, A2A protocol support for multi-agent composition, and LangGraph library version constraints. Security updates addressed vulnerabilities and improved checkpoint metadata encryption for enterprise deployments.

Enterprise ANO Adoption

LangChain announced a comprehensive integration with NVIDIA to deliver an enterprise-grade agentic AI development platform combining LangChain's LangSmith and open-source frameworks with NVIDIA Agent Toolkit. Includes Deep Agents, LangGraph for multi-agent orchestration, and AI-Q Blueprint for deep research systems.

EXL announced enhanced capabilities for its EXLerate.ai platform supported by NVIDIA AI Enterprise, including Agent Studio (no-code autonomous agent builder), AI Governance Hub, and EXLdata.ai for GPU-accelerated data processing. Enables 30-50% faster time-to-model deployment.

Meta acquired Moltbook, a social network built for AI agents to interact and share, signaling enterprise direction toward enabling autonomous AI systems to collaborate and learn from each other.

NVIDIA officially announced Agent Toolkit including NVIDIA OpenShell runtime for policy-based security and NVIDIA AI-Q Blueprint for agentic search. Built with LangChain, the platform supports 16 major enterprise software vendors and ranks #1 on deep research benchmarks.

Market analysis shows agentic AI sector expanding from $9.14B in early 2026 to over $139B by 2034. Enterprise adoption focuses on multi-step task handling and simultaneous application interaction, driving process automation across operations.

NVIDIA's 2026 report shows 64% of organizations now actively deploy AI in operations, with agentic AI adoption reaching 47-48% in telecom and retail. 88% of organizations report revenue gains from AI deployment, marking shift from pilot testing to production scaling.

ANO Security & Governance

OpenAI introduced Codex Security, an application security agent designed to identify complex vulnerabilities and minimize false positives. The agent combines agentic reasoning with automated validation to deliver high-confidence findings and actionable fixes. Available in research preview on ChatGPT Enterprise, Business and Edu customers as of March 6, 2026, with free usage offered for the first month.

CSA report documents the critical shift to proactive governance of AI agents in 2026, highlighting that autonomous agents outnumber human users by 100-to-1 ratio. Key finding: the top cloud security risk is insecure identities and machine permissions, with attackers increasingly targeting service accounts and AI agents for lateral movement.

Report from AIUC-1 Consortium shows 80% of organizations reported risky agent behaviors including unauthorized system access and improper data exposure. Only 21% of executives have complete visibility into agent permissions, tool usage, or data access patterns. Survey included CISOs from Confluent, Elastic, UiPath, and Deutsche Börse.

NIST launched the AI Agent Standards Initiative to ensure autonomous AI agents are widely adopted with confidence, function securely, and interoperate across digital ecosystems. Initiative accepts stakeholder input through RFI on AI Agent Security (deadline March 9) and will hold listening sessions beginning in April on sector-specific barriers to AI adoption.

Report from 900+ executives confirms AI agents are now production infrastructure, not experimental. Key findings: 88% of organizations reported confirmed or suspected AI agent security incidents in the last year (92.7% in healthcare); only 14.4% have full security approval for their entire agent fleet; more than half of agents operate without security oversight or logging.

Globally peer-reviewed framework identifying critical security risks for autonomous AI systems, developed through collaboration with 100+ industry experts. Provides practical, actionable guidance to help organizations secure AI agents that plan, act, and make decisions across complex workflows.

Analysis indicates 2026 is solidifying as the AI agent security market year, with vendors extending identity security foundations to provide guardrails for autonomous agents. Notes that AI agents operate as autonomous actors performing non-deterministic actions at machine speed, creating immediate security challenges for CISOs.

NIST formal Request for Information on security considerations for AI agents received 932 comments by the March 9, 2026 deadline. RFI seeks insights on practices and methodologies for measuring and improving secure development and deployment of AI agent systems.