xAI & Grok

xAI released Grok 4.20 Heavy featuring 16 specialized agents that work as a team on complex research projects. The system includes a 4-agent setup for standard users and full 16-agent capability for SuperGrok Heavy subscribers, with weekly updates planned through March.

Anthropic unveiled its enterprise agents program with customized plug-ins for specific departments. The system builds on Claude Cowork and provides private software marketplaces, controlled data flows, and IT-grade deployment controls for enterprise customers.

New Relic unveiled a no-code agentic platform for enterprises to build data observability AI agents. The platform includes prebuilt agents, supports model context protocol (MCP), and integrates with OpenTelemetry capabilities for unified data management.

Salesforce reported strong Q4 results driven by Agentforce platform growth, closing over 22,000 deals in Q4 with $1.8 billion combined ARR for Agentforce and Data Cloud. The company processed 11.14 trillion tokens, indicating production-scale deployment of AI agents.

At least nine engineers including two co-founders left xAI in early February amid company reorganization. Musk indicated the departures were strategic restructuring to improve execution speed as the company scales, rather than performance-related exits.

OpenAI Agents

AWS and OpenAI will jointly develop a Stateful Runtime Environment powered by OpenAI models through Amazon Bedrock. AWS becomes the exclusive third-party cloud distribution provider for OpenAI Frontier, OpenAI's enterprise platform for AI agents.

OpenAI hired Peter Steinberger, the Austrian developer behind OpenClaw, open-source software for building autonomous AI agents that went viral. This signals OpenAI's commitment to bringing AI agents to the masses and advancing personal agent capabilities.

OpenAI released the Codex app for MacOS, a command center for managing multiple coding agents in parallel. The app enables long-horizon tasks, background processing, and streamlined multi-agent workflows with better tracking and progress visualization.

Microsoft Copilot

Microsoft introduced Copilot Tasks, a groundbreaking AI system that works for users rather than just talking to them. This represents a shift from conversational AI to action-oriented agents that can complete tasks autonomously with their own computer and browser. The system is currently in research preview with a waitlist for broader launch.

According to Satya Nadella on X, Microsoft has added Grok 4.1 Fast to Copilot Studio's multi‑model lineup on February 19, 2026, marking a significant advancement in custom AI agent development and enhancing the platform's multi-model capabilities. This integration positions Microsoft to compete more aggressively in the AI agent market projected to reach $15.7 billion by 2028.

Copilot chat is now available directly inside apps built with Power Apps, bringing the intelligence of Microsoft 365 Copilot into the flow of business processes. This unified experience lets users ask questions, reason over in‑app data, and connect insights without leaving the application they're working in. The update also includes Power Apps Model Context Protocol (MCP) Server in public preview for better human-agent collaboration.

A core technical advance is improved grounding for agents - newly created declarative agents can now ground answers in scanned PDFs and image-based documents from SharePoint, unlocking a major class of enterprise content that was previously difficult for agents to process. The update also extends agent surfaces across Copilot app, Teams, OneDrive, Word, PowerPoint and Outlook.

Microsoft's February 2026 Teams update stitches Copilot more tightly into meetings and chat, while quietly retiring the Teams Designer bot and folding those creative capabilities into Microsoft 365 Copilot. The Designer retirement was executed with a migration path, finishing on February 27, 2026. This represents broad strategic consolidation of generative features into Copilot.

As of February 13, 2026, new benefits are available across Microsoft AI Cloud Partner Program benefits packages, including increased Azure credits, expanded Copilot benefits with more Microsoft 365 Copilot capacity packs and Copilot Studio access, and access to the new Partner Marketing Center Pro with AI-powered capabilities. Microsoft is expanding investments to accelerate organization-wide Copilot adoption with new promotional offers.

Google & Gemini

Google Cloud released its 2026 AI Agent Trends Report, forecasting that 2026 will be the year AI agents fundamentally reshape business. The report includes a companion NotebookLM and outlines five key trends defining where agents will drive new value. Google and Salesforce are building cross-platform AI agents using the Agent2Agent (A2A) protocol for interoperable agentic enterprises.

Deutsche Bank is working with Google Cloud to develop agentic AI that will monitor trading, spot anomalies in orders and trades, and flag them to human compliance officers. The bank also plans to use the technology to monitor communications of client-facing staff. Banco Santander and other banks are working with FinTech firm ThetaRay to include agentic AI in their anti-money laundering controls.

Google announced the public preview of the Developer Knowledge API with a Model Context Protocol (MCP) server, giving AI development tools a simple, machine-readable way to reach Google's official developer documentation. This fits into a broader pattern of MCP adoption across the industry, suggesting MCP is becoming the standard way to connect AI agents to live data sources.

In February 2026, Klaviyo announced a partnership with Google that connects Google's search, ads, AI, and RCS messaging with Klaviyo's real-time customer data platform to power AI‑driven, autonomous customer experiences. A distinctive element includes Klaviyo's early access to Google's Search‑to‑RCS experience. The rollout of Klaviyo's Marketing Agent and Customer Agent across the Service suite plugs directly into Google's messaging integrations.

Jules in Google AI Ultra gives 20x higher limits for intensive, multi-agent workflows at scale. Jules is Google's asynchronous coding agent, with AI Pro subscribers getting 5x higher limits versus free users. The service includes a research prototype for browser agents that lets users perform 10 tasks simultaneously for research, shopping, and booking travel.

Google is developing a UI automation framework for AI agents and assistants to intelligently execute generic tasks on users' installed apps, with the platform doing the heavy lifting so developers can get agentic reach with zero code. Starting with an early preview on Galaxy S26 series and select Pixel 10 devices, users will be able to delegate multi-step tasks to Gemini, supporting curated apps in food delivery, grocery, and rideshare categories.

Released on February 12, 2026, Google DeepMind published a comprehensive framework for Intelligent AI Delegation, designed to transform how AI agents assign, execute, and verify tasks in multi-agent systems. The release coincides with practical tools like DelegateOS, a TypeScript library that enforces delegation strictures using cryptographic tokens, suggesting the research is influencing internet infrastructure.

Agent Frameworks

LangChain and LangGraph have released their first major v1.0 versions, marking a commitment to stability with no breaking changes until 2.0. LangGraph is positioned as a lower level framework and runtime for highly custom and controllable agents, designed to support production-grade, long running agents. The v1.0 release includes durable state persistence, built-in workflow resumption capabilities, and human-in-the-loop patterns for high-stakes decision control.

MongoDB introduced the MongoDB Store for LangGraph, enabling AI agents to retain long-term memory across multiple sessions instead of only current session memory. This integration allows agents to remember and build on previous interactions, creating more intelligent, context-aware agentic systems. The platform enables use cases including customer support agents that remember preferences across channels and personal assistant applications that learn user habits.

As organizations enter 2026, they're no longer asking whether to build agents, but how to deploy them reliably at scale. Over 57% of surveyed organizations now have agents running in production environments, with another 30.4% actively developing agents with concrete deployment plans. Customer service emerged as the most common agent use case at 26.5%, with research & data analysis close behind at 24.4%, representing over half of all primary agent deployments.

Enterprise ANO Adoption

OpenAI announced major partnerships with Boston Consulting Group, McKinsey & Co., Accenture, and Capgemini to help sell and implement its new Frontier AI agent platform. The consulting firms will help clients redesign workflows, integrate AI agents with software tools and systems, and provide industry-specific expertise.

Y Combinator-backed Trace raised $3 million in seed funding to address AI agent adoption challenges through workflow orchestration and context mapping. The company competes with Anthropic's newly launched enterprise agents focused on pre-built departmental plug-ins.

Gartner predicts 40% of enterprise applications will embed task-specific AI agents by end of 2026, up from less than 5% in 2025. The report highlights multi-agent orchestration patterns and governance challenges as key implementation concerns.

Tabnine launched its Enterprise Context Engine to ground AI assistance in enterprise context, enabling safer automation and more accurate software development. The platform addresses the challenge of AI agents understanding enterprise system structures and dependencies.

Salesforce closed over 22,000 Agentforce deals in Q4, representing nearly 50% quarter-over-quarter growth. Combined annual recurring revenue for Agentforce and Data Cloud reached approximately $1.8 billion, up from $1.4 billion three months prior.

BigBear.ai completed its $250 million acquisition of Ask Sage, a generative AI platform for government and regulated industries. The acquisition expands BigBear.ai's presence across defense, intelligence, and national security markets with FedRAMP High authorized AI tools.

ANO Security & Governance

NIST's Center for AI Standards and Innovation announced a major initiative to foster industry-led AI agent standards and protocols. The initiative focuses on three pillars: facilitating industry standards development, fostering open source protocols, and advancing research in AI agent security and identity to enable trusted adoption across economic sectors.

A survey of 919 executives reveals that 80.9% have moved AI agents into production, but only 47.1% are actively monitored or secured. The report shows 88% of organizations experienced confirmed or suspected AI agent security incidents in the past year, with healthcare organizations reporting 92.7% incident rates.

OpenAI's Frontier platform allows enterprises to treat AI agents like human employees with onboarding processes and feedback loops. The open platform can manage agents built outside of OpenAI, with early customers including Intuit, Uber, State Farm, and Thermo Fisher reporting significant time and cost savings.

Microsoft issued security warnings about OpenClaw, highlighting that self-hosted agent runtimes should be treated as untrusted code execution with persistent credentials. The post emphasizes the need for isolated environments, dedicated credentials, and continuous monitoring as agent hijacking becomes a real threat.

Analysis of the shift from AI tools to AI collaborators, where prompt injection attacks can turn trusted AI agents into malicious insider threats. The article predicts this threat will significantly increase in 2026 as enterprises grant more autonomy to AI agents with access to internal systems.

Reports show multi-turn attacks achieving 92% success rates across open-weight models, with compromised agents able to execute unauthorized commands and move laterally across systems. The AI supply chain presents additional risks with malicious code embedded in model objects that can trigger automatically during initialization.

Anthropic unveiled its enterprise agents program with Claude Cowork, featuring private software marketplaces, controlled data flows, and customized plug-ins. The system allows corporate IT departments to deploy Claude-powered agents with the same controls expected for enterprise software.

ANO Academic Research

MiniMax announced MaxClaw, an open-source agent system featuring one-click deployment and native long-term memory capabilities. The platform enables persistent AI workflows with built-in skills including image generation, web search, and multi-agent research teams.

Multi-agent workflows on the Databricks platform saw 327% growth, with the Supervisor Agent being the most popular for directing other agents. The company introduced 'lakebase' architecture specifically designed for multi-agent database operations.

NVIDIA released Nemotron 3 open models with breakthrough hybrid mixture-of-experts architecture designed specifically for multi-agent systems at scale. Nemotron 3 Nano delivers 4x higher throughput than previous versions for multi-agent deployments.

Anthropic unveiled its enterprise agents program with private software marketplaces and customized plug-ins for Claude Cowork. The system allows corporate IT departments to deploy Claude-powered agents with centralized controls and tailored workflows.

Analysis shows most multi-agent AI systems fail due to memory coordination issues where agents reexecute tasks and produce inconsistent results. The article details how Agent A completes work, Agent B repeats it unknowingly, and Agent C reconciles conflicting outputs.

ANO Regulation & Policy

NIST's Center for AI Standards and Innovation launched the AI Agent Standards Initiative to ensure AI agents can function securely and interoperate across the digital ecosystem. The initiative focuses on industry-led standards development, open source protocols, and AI agent security research.

OpenAI's Frontier platform allows enterprises to build and deploy agents while managing them like human employees. Early customers include Intuit, Uber, State Farm and Thermo Fisher, with one financial services firm reporting 90% more time back for client-facing teams.

Anthropic unveiled its enterprise agents program with customized plug-ins for specific departments. The program focuses on making Claude-powered agents easier to deploy within companies with private software marketplaces and controlled data flows.

New Relic unveiled a no-code agentic platform that lets enterprises build data observability AI agents. The platform supports the Model Context Protocol (MCP) and enables companies to deploy prebuilt agents while managing existing bots.

Red Hat introduced AI Enterprise, integrating high-performance AI inferencing, model tuning, and agent deployment capabilities. The platform runs on Red Hat OpenShift and includes collaboration with Nvidia for accelerated enterprise AI development.

Salesforce reported strong Q4 results driven by Agentforce platform growth, closing over 22,000 deals with combined ARR of $1.8 billion for Agentforce and Data Cloud. The platform served 11.14 trillion tokens, indicating production-scale deployment.

Sweep announced the Multi-Org Agent for Salesforce, designed for enterprises with multiple Salesforce organizations. The agent provides cross-system intelligence and coordination capabilities for complex enterprise environments.