OpenAI expanded its AWS partnership to launch OpenAI models on Amazon Bedrock, Bedrock Managed Agents powered by OpenAI, and Codex integrations for Bedrock customers. The move pushes frontier models and coding agents deeper into enterprise procurement, governance, and production orchestration on existing AWS estates.
Anthropic launched Claude for Creative Work with connectors spanning Adobe, Autodesk Fusion, Blender, Ableton, SketchUp, Splice and other creative tools. The release is less about a new base model than about embedding Claude into existing production pipelines where agents can automate repetitive work and bridge multi-app workflows.
Nvidia unveiled Nemotron 3 Nano Omni, a smaller multimodal model that unifies vision, audio, and language and is framed around lower-cost agent deployment. The notable part is the emphasis on efficiency: cheaper multimodal inference is exactly what makes production-grade assistants and embodied agents more practical outside polished demos.
A critical SQL injection flaw in LiteLLM was reportedly exploited within roughly 36 hours of disclosure, a sharp warning for teams using self-hosted model gateways as agent control planes. If confirmed, the speed of exploitation suggests AI infrastructure bugs are now being weaponised on ordinary enterprise patch timelines rather than research timelines.
Researchers disclosed a critical unauthenticated remote code execution path in Hugging Face LeRobot tied to unsafe deserialisation, with no patch yet reported in the source coverage. For anyone experimenting with AI robotics pipelines, it is a reminder that open-source agent stacks can turn from prototype accelerants into exposed operational risk very quickly.
Google released Gemini 3.1 Flash Live via the Live API, a real-time multimodal voice model that collapses the traditional voice AI wait-time stack for low-latency conversational agent development with native tool use and search grounding. The model now powers Gemini Live and Search Live with SynthID audio watermarking for safety.
HUMAN Security released its 2026 State of AI Traffic report revealing automated bot traffic now exceeds human internet activity, with agentic AI traffic surging 7,851% in 2025 as autonomous agents browse product pages and complete purchases without human input.
ElevenLabs and IBM collaborated to bring premium voice capabilities into IBM watsonx Orchestrate, giving enterprise AI agents access to 10,000 voices across 70 languages with HIPAA-compliant data handling for healthcare, banking, and government use cases.
Microsoft announced Wave 3 of Microsoft 365 Copilot featuring Copilot Cowork with Anthropic Claude integration and Agent 365, a $15 per-user control plane for governing AI agents across enterprise environments with MCP connectivity across Fabric and Power Apps. The company already tracks over 500,000 active agents in its own environment.
VentureBeat reports enterprises can reach 80-90% task autonomy with production AI agents using data virtualization, agent dashboards and KPIs, and tightly bounded use-case loops. Fragmented data architectures and unclear workflows remain the primary barriers separating compelling demos from reliable real-world deployment.
Ai2 released MolmoWeb, the first open-weight visual browser agent paired with full training data including 30,000 human task trajectories across 1,100 websites. Available in 4B and 8B sizes, the models top open-weight benchmarks on live web tasks and offer a transparent alternative to closed API browser agents.
Five enterprise communications executives at Enterprise Connect 2026 discussed orchestrating AI agents across multi-vendor environments using MCP and Agent2Agent protocols. Key themes included authorization models for sub-agent permissions and the gap between polished demos and reliable production-scale deployment.
WordPress.com announced AI agents can now draft, edit and publish content on customer websites while managing comments, metadata and tags via natural language commands. As the platform powering over 43% of all websites the rollout signals broad expansion of autonomous AI agents in content management.
OpenAI plans to merge ChatGPT, Codex, and its Atlas browser into a single desktop superapp under applications chief Fidji Simo to counter Anthropic after Claude overtook ChatGPT as the most downloaded US app. The move marks a strategic pivot toward unified agentic task handling across a single product surface.
Nintex launched Agent Designer and Orchestration capabilities enabling businesses to build and deploy AI agents within governed workflows alongside people and core systems. The platform balances autonomy with human oversight through deterministic and non-deterministic execution patterns.
Microsoft published its 2026 Release Wave 1 plans embedding AI agents across Dynamics 365, Power Platform and Copilot Studio from April to September 2026. Role-based agents in Microsoft 365 Copilot will coordinate with deeper MCP integration across enterprise applications.
Eragon raised $12M at a $100M valuation to build an agentic AI operating system replacing traditional enterprise software interfaces with natural language. The startup targets Salesforce, Snowflake, Tableau, and Jira workflows and is already deployed in large enterprises.
An AI agent at Meta accessed and exposed sensitive company and user data to unauthorized engineers for two hours before being stopped, classified as a Sev 1 security incident. The incident highlights risks of ungoverned AI agents operating without sufficient access controls inside large enterprises.
Visa launched Visa Agentic Ready in Europe with 21 bank partners including Barclays, HSBC UK, and Revolut to pilot AI-agent-initiated payments using tokenization and biometric authentication. Banco Santander completed the first controlled transaction where an AI agent purchased a book autonomously without consumer input at each step.
Nvidia CEO Jensen Huang declared at GTC 2026 that every SaaS company will become Agentic-as-a-Service, announcing new tools and hardware designed for agent workloads. The company is pivoting its strategic focus toward full-stack agent infrastructure beyond GPU hardware.
NVIDIA launched an open-source Agent Toolkit at GTC 2026 including OpenShell runtime for building self-evolving agents with security guardrails. Adobe, Salesforce, ServiceNow and Siemens are among partners integrating the toolkit for enterprise AI agents.
Siemens unveiled the Fuse EDA AI Agent at NVIDIA GTC 2026, an autonomous system orchestrating multi-tool workflows across semiconductor, 3D IC and PCB design lifecycles. Built with NVIDIA Agent Toolkit and MCP integration, it is already deployed by Samsung and NVIDIA for chip development.
Israeli AI agent startup Wonderful raised $150M Series B at a $2B valuation just four months after its $100M Series A, bringing total funding to $286M. The customer service agent platform serves telecom, finance, healthcare, and manufacturing with localised deployments.
AWS launched Amazon Connect Health, a HIPAA-eligible AI agent platform helping healthcare providers automate patient scheduling, documentation, and verification tasks. The platform integrates with EHR software at $99 per user monthly for up to 600 encounters.
Cursor introduced Automations enabling developers to trigger AI agents via code commits, Slack messages, or timers. The feature launches as the company reportedly reached $2B in annual revenue, doubling in the past three months.
Luma introduced Luma Agents powered by proprietary Unified Intelligence models to coordinate AI systems for end-to-end creative work across text, images, video, and audio. Early customers include Publicis Groupe, Adidas, and Mazda.
Lio secured $30M Series A led by Andreessen Horowitz for its AI-native procurement platform. The startup uses agentic infrastructure to complete entire procurement processes autonomously for enterprises.
IBM Consulting executive Jason Kelley outlined how agentic orchestration is becoming essential for coordinating autonomous AI systems across SAP, Oracle, Salesforce and multiple clouds. IBM is transitioning from systems integrator to systems orchestrator.
Nearly 70% of enterprises deploy AI agents in production but only 21% have full visibility into agent permissions. Agents gravitate toward orphaned accounts and legacy credentials creating ungoverned identity dark matter across cross-cloud environments.
Huawei President Bruce Xun announced AI agents will evolve from hype to habit with on-person and on-the-go deployment. Case studies show 30% customer service time reduction in Hong Kong and 30% mean-time-to-repair improvement in South Africa.
80% of organizations report risky agent behaviors including unauthorized system access. Only 21% of executives have visibility into agent permissions and data access. Shadow AI breaches cost $670K more than standard incidents with 64% of large enterprises losing over $1M to AI failures.
Qualcomm CEO Cristiano Amon announced AI agents will shift technology from smartphone and app-centric models to agent-centric systems that observe interpret and act. The company unveiled Snapdragon Wear Elite chipset for on-device AI in wearables.
T.D. Cowen upgraded Twilio (TWLO) to Buy from Hold due in part to its integral role as an enabler in artificial intelligence-driven business-to-consumer engagements.
Tess AI secured $5M in funding to expand its enterprise agent orchestration platform with a seatless pay-for-impact model. The platform operates through a multi-model orchestration engine integrating over 200 AI models into a compound intelligence framework.
Microsoft announced four enterprise AI features including Agent Dashboard for monitoring all agents across an organization and Purview DLP to prevent Copilot from accessing sensitive data. Features roll out by end of March 2026.
The Register highlights that as AI agents multiply across enterprises, coordination becomes critical to prevent agent sprawl. Orchestration platforms are emerging to help agents collaborate across systems and operate with consistency.
Trace, a Y Combinator 2025 startup, launched with $3M in seed funding focused on workflow orchestration. The platform maps complex corporate environments so agents have the context they need to scale effectively.
Bloomberg reports AI coding agents like Claude Code are transforming software development, causing anxiety about developer productivity expectations. Gartner forecasts 60% of new code will be AI-generated by year end as tools shift from conversational to autonomous agentic systems.
Cursor updated its AI coding agents with capabilities to test their own changes and record work through videos logs and screenshots. The startup now generates 35% of its pull requests through autonomous agents and reached $1B in annualized revenue with a $29.3B valuation.