xAI & Grok

xAI released Grok 4.20 Heavy featuring 16 specialized agents that work together as a team on complex research projects. The system allows SuperGrok Heavy tier users to access full multi-agent capabilities, while standard users get a four-agent setup for balanced reasoning outputs.

xAI launched Grok 4.20 with multi-agent reasoning capabilities, including a live thinking interface where users can watch agents collaborate step-by-step. The system is available to all users including free accounts, making agent capabilities widely accessible.

xAI announced major updates including Grok Voice Agent API with multiple expressive voices, enhanced enterprise features with Vault security, and expanded real-time search capabilities. The company also completed a $20 billion Series E funding round exceeding initial targets.

At least nine engineers including two co-founders have exited xAI amid company reorganization. Musk indicated the departures were strategic restructuring decisions rather than voluntary exits, as the company scales and evolves its structure.

Weekly briefing covering xAI restructuring following co-founder exits and launch of 'macrohard' to let agents run companies. Also mentions Anthropic's collaborative agents launch and Goldman Sachs partnering with Anthropic for AI agent banking automation.

OpenAI Agents

OpenAI announced partnerships with major consulting firms including Boston Consulting Group, McKinsey & Co., Accenture, and Capgemini to help sell and implement the Frontier AI agent platform. These Frontier Alliances partnerships involve dedicated practice groups and certified teams to help enterprise clients deploy agents at scale.

OpenAI launched Frontier as an end-to-end platform for enterprises to build and manage AI agents, including support for agents built outside of OpenAI. Gartner called agent management platforms the 'most valuable real estate in AI' and necessary infrastructure for enterprises to adopt AI.

OpenAI released the Codex app for MacOS, a command center for managing multiple coding agents in parallel with long-horizon and background tasks. ChatGPT also received Agent Mode improvements, reducing 'laziness' and making it more persistent on repetitive and tedious tasks.

Anthropic & Claude

GitHub expanded access to Claude and Codex coding agents for Copilot Business and Pro customers. Users can run these agents directly within github.com, GitHub Mobile, and VS Code with unified governance and shared context across platforms.

Anthropic unveiled its enterprise agents program with pre-built agents for finance, legal, and HR departments. The system includes private software marketplaces, controlled data flows, and customized plug-ins for corporate deployment.

OpenAI announced partnerships with major consulting firms to sell and implement its Frontier AI agent platform. The partnerships pose a direct threat to traditional SaaS vendors by offering alternative agent orchestration solutions.

AWS introduced Claude Sonnet 4.6 in Amazon Bedrock and new open-source Agent Plugins that extend coding agents with AWS deployment capabilities. The plugins enable architecture recommendations and infrastructure-as-code generation directly from coding agents.

Anthropic partnered with Infosys to develop industry-specific AI agents for telecom, financial services, and manufacturing. The collaboration integrates Claude models with Infosys Topaz to help regulated industries adopt AI with proper governance.

Samsung expanded its Galaxy AI ecosystem to include Perplexity as an additional AI agent alongside existing services. The multi-agent system operates at the framework level with deep OS integration for seamless user experiences.

Microsoft Copilot

Microsoft introduces Power Apps Model Context Protocol (MCP) Server and enhanced agent feed into public preview. New updates to computer-using agents improve UI automation with secure credentials, detailed monitoring, and scalable Cloud PC capacity.

Computer-using agents in Microsoft Copilot Studio now support multiple foundation models including Anthropic's Claude Sonnet 4.5 alongside OpenAI's Computer-Using Agent. Features improved security, governance, and scalability for enterprise deployments.

Microsoft 365 Copilot declarative agents now use the GPT-5.1 model with automatic model selection for performance and reasoning. Users can scope grounding for declarative agents to specific data sources for more precise responses.

Microsoft outlines organizational capabilities needed for agent adoption at scale in 2026. Conversation became the agent-making interface in both Copilot Studio and Agent Builder, enabling natural language agent creation without technical barriers.

Copilot Chat model selector now offers GPT-5.2 for faster responses or deeper reasoning. Admins can set up connector authentication faster through simplified OAuth experience and manage Agent Blueprints through Developer Portal.

Microsoft launches new agent architecture guidance hub and refreshed Copilot Studio guidance hub. Resources provide end-to-end documentation across agent lifecycle from design through operations and governance.

Google & Gemini

Google Cloud released its 2026 AI Agent Trends Report, forecasting that AI agents will fundamentally reshape business through multi-agent workflows, agentic customer service, and security automation. The report highlights Google's partnership with Salesforce on the Agent2Agent (A2A) protocol for cross-platform AI agent interoperability.

Google announced the Universal Commerce Protocol (UCP) for AI agent-based shopping, allowing merchants to offer discounts directly in AI mode results. The protocol enables checkout directly from U.S.-based retailers while researching products, with support for Google Pay and future PayPal integration.

Google published its 2026 Responsible AI Progress Report detailing how the company is restructuring governance for agentic AI, browser agents, and AGI preparation. The report reveals a seven-layer governance framework and emphasizes the shift from controlling model outputs to governing autonomous agent actions.

Google introduced new AI tier structures including Jules (asynchronous coding agent), multi-agent workflow capabilities in Ultra tier, and browser agents for research tasks. The Ultra plan provides 20x higher limits for intensive multi-agent workflows at scale with the highest agent request capabilities.

Google's Ads Decoded podcast revealed how AI Mode and Direct Offers are reshaping search ad experiences, with shopping ads now integrated into AI-generated responses. The Universal Commerce Protocol launch allows retailers to present exclusive discounts directly within AI Mode conversations.

Google Cloud released a specialized report on AI agent trends for retail and CPG industries, exploring five key trends including agents for every employee and upskilling talent as the ultimate driver of business value. The report emphasizes cultural change and questioning old assumptions to unlock AI agent value.

Agent Frameworks

LangChain and LangGraph have officially released their v1.0 versions, marking the first stable major releases in the agent framework space. The releases include enhanced stability commitments, new middleware capabilities for LangChain, and production-ready features for LangGraph including durable state and human-in-the-loop patterns.

MongoDB introduced the MongoDB Store for LangGraph, enabling AI agents to retain cross-session memory and learn from previous interactions. This integration provides both short-term and long-term memory solutions for production-ready agent deployments.

LangChain's 2026 State of AI Agents report reveals that 57.3% of surveyed organizations now have agents in production, with customer service (26.5%) and research & data analysis (24.4%) as the most common use cases. The report shows significant growth from 2025 and highlights the shift from 'if' to 'how' and 'when' for agent deployment.

Anthropic's research from 500+ technical leaders shows 80% of enterprises already report measurable ROI from AI agents in 2026. The report identifies integration challenges, data quality issues, and change management as primary hurdles, while 90% of leaders report agents are reshaping how teams work.

Enterprise ANO Adoption

Salesforce closed over 22,000 Agentforce deals in Q4, representing nearly 50% quarter-over-quarter growth. Combined ARR for Agentforce and Data Cloud reached $1.8 billion, up from $1.4 billion in Q3, with the platform serving 11.14 trillion tokens in the quarter.

Capgemini partnered with OpenAI to help enterprises move from AI pilots to scaled deployments using the Frontier platform. The goal is to help enterprises design, deploy, and run AI agents across multiple business units rather than as isolated pilots.

Multi-agent workflows grew more than 300% over several months as organizations moved from pilot phases into production. Capital One built multi-agent workflows embedded directly into operational systems, with 43% of CFOs saying agentic AI could have high impact on budget planning.

Tabnine launched its Enterprise Context Engine to solve the lack of organizational context for enterprise AI adoption. The platform builds a continuously evolving model of an organization's software systems and documentation, enabling AI agents to reason about how systems work.

NIST's Center for AI Standards and Innovation launched the AI Agent Standards Initiative on February 17, 2026. The initiative aims to ensure AI agents capable of autonomous actions can function securely and interoperate across digital ecosystems while fostering U.S. technological dominance.

ANO Security & Governance

NIST's Center for AI Standards and Innovation launched the AI Agent Standards Initiative on February 17, 2026 to ensure secure and interoperable AI agent adoption. The initiative focuses on developing industry-led standards, fostering open source protocols, and advancing AI agent security research.

NIST launched the AI Agent Standards Initiative seeking feedback on secure use of fast-evolving agentic AI technology. The project aims to build public trust and prevent AI agents from 'wreaking havoc' as they become more capable and autonomous.

Survey of 919 executives reveals structural gaps in AI agent security, with 88% of organizations reporting AI agent security incidents. Only 47.1% of AI agents are actively monitored, creating significant 'Shadow AI' risks in enterprise deployments.

Enterprise AI security risks are rising as agentic AI deployments expand attack surfaces. Multi-turn attacks achieved 92% success rates against open models, while compromised agents can execute unauthorized commands and move laterally across systems.

Veza launched AI Access Agents on February 25, 2026 to automate identity governance tasks and enhance AI agent security capabilities. The platform addresses identity blind spots created by semi-autonomous AI entities and expanded MCP server discovery.

Analysis of AI agent network security gaps following NIST's initiative launch. Dark Reading poll found 48% of security professionals rank agentic AI as the top attack vector for 2026, highlighting need for identity-based microsegmentation.