xAI & Grok

xAI released Grok 4.20 Heavy featuring 16 specialized agents that work as a team on complex research projects. The standard version uses 4 agents while the Heavy version for SuperGrok Heavy subscribers provides access to all 16 agents for demanding workloads.

xAI launched Grok 4.20 with a multi-agent system featuring Grok, Harper, Benjamin, and Lucas working together on every request. Users can watch the agents reason step-by-step through a new live thinking interface, available to both free and premium users.

xAI is testing new features allowing up to 8 coding agents to work simultaneously on tasks, plus Arena Mode where agents compete on problems with the strongest solution winning. These features are in active testing for the next Grok Build release.

OpenAI announced Frontier, an end-to-end platform for enterprises to build and manage AI agents. The open platform allows management of agents built outside OpenAI and includes onboarding processes and feedback loops similar to managing human employees.

Snowflake and OpenAI announced a $200 million partnership on February 2, 2026, to enable global enterprises to unlock greater value from proprietary data with AI. The collaboration focuses on bringing enterprise-ready AI capabilities to Snowflake's data platform.

At least nine engineers including two co-founders announced exits from xAI in early February 2026. Musk addressed the departures at an all-hands meeting, suggesting they were organizational restructuring decisions rather than voluntary departures as xAI scales.

OpenAI Agents

OpenAI announced Frontier as an end-to-end platform for enterprises to build and manage AI agents. It's an open platform allowing users to manage agents built outside of OpenAI, with notable enterprise deals including ServiceNow and Snowflake.

OpenAI hired Peter Steinberger, the Austrian developer behind OpenClaw—viral open-source software for building autonomous AI agents. The move signals OpenAI's aggressive bet on the future of AI being about what models can do rather than what they can say.

OpenAI Frontier addresses system fragmentation by providing shared business context, agent onboarding, and identity governance for enterprise AI agent deployment. The platform emphasizes reliable, scalable integration into real company systems and workflows.

Anthropic partnered with Infosys on February 17, 2026, to develop enterprise AI solutions across telecommunications, financial services, and manufacturing. The collaboration focuses on agentic AI systems that independently handle multi-step tasks like processing claims and managing compliance.

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 in defense, intelligence, and national security markets with AI agents and orchestration capabilities.

Anthropic & Claude

Anthropic's official announcement of Claude Sonnet 4.6 featuring a 1M token context window in beta, significant improvements in agentic workloads, and strong performance on complex code fixes. The model demonstrates 94% accuracy on insurance benchmarks for computer use tasks.

Anthropic unveiled Claude Code Security in limited research preview for Enterprise and Team customers. The capability scans codebases for security vulnerabilities and suggests targeted software patches, designed to help defenders counter AI-enabled attacks by adversaries.

Strategic collaboration announced to develop enterprise AI solutions across telecommunications, financial services, manufacturing, and software development. Partnership will establish a dedicated Anthropic Center of Excellence and focus on agentic AI systems for complex multi-step tasks.

OpenAI launched Frontier, an end-to-end platform for enterprises to build and manage AI agents. The platform functions like employee management systems, offering agent onboarding, feedback loops, and the ability to manage agents from other vendors beyond OpenAI.

Analysis of Anthropic's release of Claude Opus 4.6 with autonomous agent teams capability and the controversy over military use through Palantir partnership. The model can coordinate multiple AIs working in parallel and navigate web applications with human-level capability.

Anthropic released Claude Opus 4.6 designed to broaden capabilities beyond software development to knowledge work. The model has evolved to be useful for product managers, financial analysts, and various industries, not just professional software developers.

Google & Gemini

Google Cloud released its 2026 AI Agent Trends Report forecasting that AI agents will fundamentally reshape business. The report details how employees will delegate tasks to different AI agents, shifting work from routine execution to strategic direction, and how multiple agents will collaborate through the Agent2Agent protocol.

Google published its latest Responsible AI Progress Report detailing how the company is restructuring governance for agentic AI systems. The report covers a seven-layer governance framework spanning research, policies, testing, mitigation, launch review, monitoring, and governance for AI agents that can act autonomously.

Analysis of Google's February 2026 Responsible AI Progress Report reveals how the company is rebuilding governance structures around agentic AI. The report shows Google now operates a framework covering emergent behaviors from agent-to-agent interaction and collaboration with UK's AI Safety Institute.

Google introduced Gemini 3.1 Pro with more than double the reasoning performance for developing advanced agents and enterprise applications. The model emphasizes long-horizon planning and is available through Vertex AI, targeting developers building autonomous agent systems.

Google and Southeast Asia's Sea Ltd formed a strategic partnership to develop AI tools including an agentic shopping prototype embedded within Shopee's marketplace. The collaboration extends AI agents into task execution across commerce workflows and game development productivity.

Google shipped major Deep Think updates powered by Gemini 3.1 Pro, which scored 77.1% on ARC-AGI-2 benchmark. The release includes animated SVG generation from text prompts and is rolling out across Gemini app, NotebookLM, Google AI Studio, and Vertex AI for agent development.

Google launched the open Agent2Agent (A2A) protocol with support from over 50 technology partners including Salesforce, SAP, ServiceNow, and others. The protocol allows AI agents to communicate with each other, securely exchange information, and coordinate actions across enterprise platforms.

Google and Salesforce announced the Universal Commerce Protocol (UCP) at NRF 2026, co-developed with major retailers including Shopify, Walmart, Target, and over 20 ecosystem partners. UCP standardizes AI agent-powered shopping experiences where agents execute everything from product discovery to purchase on behalf of consumers.

Agent Frameworks

OpenAI launched Frontier, a new platform designed for enterprises to build and deploy agents while treating them like human employees. On Thursday, AI giant OpenAI announced the launch of OpenAI Frontier, an end-to-end platform designed for enterprises to build and manage AI agents. Frontier users can program AI agents to connect to external data and applications, which allows them to execute tasks far outside of the OpenAI platform. Users can also limit and manage what these agents have access to, and what they can do.

OpenAI has introduced Frontier, an enterprise platform designed to operationalize AI agents as digital coworkers across business systems. The move highlights a widening gap between AI model potential and enterprise execution, and intensifies competition to define the agent platform layer. Dozens of companies have already adopted Frontier, including Intuit, Uber, State Farm and Thermo Fisher. A global financial services firm using the technology got 90% more time back for their client-facing team.

We're releasing LangChain 1.0 and LangGraph 1.0 — our first major versions of our open source frameworks! LangGraph is a lower level framework and runtime, useful for highly custom and controllable agents, designed to support production-grade, long running agents. This is the first stable major release in the durable agent framework space — a major milestone for production-ready AI systems. After more than a year of iteration and widespread adoption by companies like Uber, LinkedIn, and Klarna, LangGraph is officially v1.

The collaboration integrates Anthropic's Claude models and Claude Code with Infosys Topaz, an AI-first set of services, solutions, and platforms using generative and agentic AI technologies, to help companies speed up software development and adopt AI with the governance and transparency that regulated industries require. A core focus of the collaboration is agentic AI—systems that go beyond answering questions to independently handling multi-step tasks like processing claims, generating and testing code, or managing compliance reviews. Using tools like the Claude Agent SDK, Infosys and Anthropic will help clients build AI agents that work persistently across long, complex tasks and processes.

2026-02-18 - agent-server - Fixed a regression in assistant creation by ensuring metadata and config are populated as empty objects {} instead of null when not provided. Added stricter version constraints for langgraph libraries in executor Docker images to prevent unintended upgrades. Updated @langchain/langgraph from version 1.1.0 to 1.1.2, introducing mixed schema support for StateGraph and type bag patterns for GraphNode and ConditionalEdgeRouter utilities.

Enterprise ANO Adoption

OpenAI introduced Frontier, a comprehensive enterprise platform for building, deploying, and managing AI agents that can do real work. The platform addresses system fragmentation by providing shared business context, onboarding capabilities, and identity governance for agents across organizations.

Oracle announced new AI agents within Fusion Cloud Applications at Oracle AI World Mumbai on Feb 10, 2026. The agents include Planning Cycle Agent for automating task coordination and Component Replacement Agent for minimizing disruptions, embedded within supply chain processes.

Multi-agent workflows grew more than 300% over several months according to Databricks. Organizations are deploying coordinated digital networks where agents handle different workflow stages - data harvesting, validation, transaction execution, and compliance.

Anthropic released Claude Opus 4.6 with 1-million-token context and "agent teams" for dividing complex tasks. OpenAI launched GPT-5.3-Codex with 25% faster performance and OpenAI Frontier enterprise platform simultaneously.

Cogent Security raised $42 million to accelerate AI agent development for vulnerability remediation at enterprise scale. Founded in 2025, the company develops autonomous AI agents that operate across security and engineering systems while working with dozens of Fortune 1000 companies.

Oracle announced new role-based AI agents for CX at Oracle AI World Mumbai. The agents are embedded within marketing, sales, and service processes, including Program Planning Agent that helps marketers plan cross-sell and up-sell programs by defining goals, audience, and campaign narratives.

ANO Security & Governance

NIST launched the AI Agent Standards Initiative on February 17, 2026, to foster industry-led technical standards for AI agent security and interoperability. The initiative focuses on advancing research in AI agent security, identity protocols, and establishing trusted adoption across economic sectors.

Survey of 919 executives reveals 88% of organizations report confirmed or suspected AI agent security incidents in the last year. Only 47.1% of AI agents are actively monitored, with 14.4% having full security approval for their entire agent fleet.

OpenAI's Frontier platform enables enterprises to build and deploy agents with external data connections and access controls. The open platform manages agents built outside OpenAI, positioning enterprise adoption as a main focus area for 2026.

Cisco's State of AI Security 2026 report details materialized AI vulnerabilities and exploits from late 2025. The report examines prompt injection evolution, AI supply chain fragility, and Model Context Protocol agentic AI risks as adversaries use agents for attack campaigns.

Analysis of the shift from AI as a 'tool' to 'collaborator' with autonomous agents executing tasks. Highlights prompt injection risks turning trusted AI agents into insider threats when they access internal data systems like OneDrive and Salesforce.

Analysis shows multi-turn attacks against AI agents achieve 92% success rates across eight open weight models. Organizations granted agentic systems authority to execute tasks and access databases, creating new exposure across model interfaces and supply chains.