AI agents

At Microsoft Build, Microsoft announced Microsoft Execution Container (MXC), a runtime containment offering for agentic AI workloads, and updates to its multi-agent vulnerability research system MDASH. MXC is a policy-driven execution workflow that lets developers specify what an AI agent can access, such as files, networks, resources and credentials, and enforces those boundaries at runtime. OpenClaw and NVIDIA's OpenShell are already adopting MXC.

The source argues that high-autonomy AI agents with broad permissions and unfettered access create serious risk. It urges enterprises to act now before such agents cause a damaging incident.

Microsoft announced Project Solara at Build 2026, a new operating system built on Android rather than Windows, designed for gadgets that run AI agents. The company demonstrated two concept devices: a desk device resembling an Amazon Echo Show that unlocks with facial recognition, and a wearable badge with a camera and fingerprint scanner that can wake an AI agent.

AWS describes how SaaS providers can use resource-based policies on Amazon Bedrock AgentCore to control access to a shared AgentCore Runtime and its endpoint per tenant. The post walks through a multi-tenant customer service platform where one tenant gets cross-account access from its own AWS account, while another is restricted to invocations from a private VPC for regulatory compliance.

Google has launched Spark, a new always-on AI agent positioned as its agentic answer to general-purpose assistance. The source text is an excerpt from a Verge review that describes the reviewer's trip-planning experience with the product, but the excerpt cuts off before the full assessment.

KPMG reports that 68% of health-care providers have adopted AI agents into their workforce. At Hospital for Special Surgery, AI agents process insurance claims at 1,100 per month, cutting the appeals stage from 45 minutes to five and raising the appeal success rate from 65% to 100% in nine months. HSS is also deploying an AI scheduling and triage service, built with Ema Unlimited, that escalates sensitive or uncertain cases to human specialists.

Google's Gemini Spark, an AI agent that Google says can run multi-step tasks in the background, was tested by a reviewer who found it impressively capable. The reviewer questions whether it is worth its financial cost and potential privacy tradeoffs.

The Linux Foundation is inviting contributions to DNS-AID, a proposed standard that lets AI agents and MCP servers discover, verify and communicate with one another using existing DNS infrastructure. The proposal suggests domain owners publish a well-known address, _index._agents.{domain}, as a starting point for agent discovery. DNS-AID was initially developed at Infoblox, with contributions from Deutsche Telekom and Amazon in the latest internet draft.

Adobe's Firefly AI Assistant, currently in beta, is a conversational agent that operates Adobe's design apps on a user's behalf to handle routine work while keeping creative control with the user. The Verge's reviewer says the assistant explains its editing steps well but is not impressed by the results.

Okta beat Wall Street's fiscal first-quarter estimates, reporting adjusted earnings of 91 cents per share against 85 cents expected and revenue of $765 million against $752 million expected. Revenue grew 11% year over year, and shares gained 8%. CEO Todd McKinnon said demand for identity tools is rising with agentic AI, though AI is not yet a majority of revenue.

The article argues that agentic AI is not inherently risky and that the danger comes from how organizations deploy it. It describes agents as models interacting with software tools, and says the risk lies where those two overlap.

Endava, a global software contracting firm, has adopted Codex across its engineering teams and now describes itself as an agentic organization, where senior expertise is codified into agents that support work from intake through delivery. Executives Joe Dunleavy and Mike Krolnik say Codex lets small teams deliver more quickly and lets junior engineers produce more senior-level output. In one engagement, a two-hour recorded meeting was turned into a working requirements specification in about two one-hour sessions instead of weeks of revision.

Organizations adopting AI are urged to pair that investment with a cybersecurity strategy, as GenAI platforms have moved from pattern-matching LLMs to tool-calling agents. The source argues that agentic attacks move faster than manual human defenses, and that AI agents connected to many systems create a flat network at odds with segmentation principles. In Armis' 2026 State of Cyberwarfare Report, 43% of respondents said their organizations still detect and respond to significant attacks as they happen or after they occur.

Robinhood is opening its trading platform to AI agents, according to an announcement on Wednesday. Traders can create a separate account for an agent and fund it with a specific amount, letting the agent buy and sell stocks across the market. Robinhood warns that agentic trading involves significant risk, including possible loss of the entire investment.

Warp, a terminal company, is building its open-source workflows around GPT-5.5 and a model it calls Open Agentic Development, in which agents plan work, write code, test changes and open pull requests while humans set objectives and supervise outcomes. According to the source, GPT-5.5 used 30% fewer tokens per agentic coding task than GPT-5.4 in Warp's internal benchmarks, and agents now co-create around 90% of Warp's pull requests. The company's orchestration platform, Oz, coordinates agents across local and cloud environments.

Every company needs an agentic AI strategy, according to the source. The tools that would let agentic AI frameworks be adopted safely and securely are only just beginning to appear.

Organizations are layering AI agents onto existing operations rather than redesigning their operating models, according to PwC UK Consulting's Prasun Shah. Ema, an enterprise agentic AI platform, has coined the term agentic business transformation (ABT), which rests on three pillars: technology stack, workforce and success metrics. The article reports that 85% of organizations want to be agentic within three years, but 76% say their operations and infrastructure cannot yet support that change.

Researchers behind a paper published this month argue that enterprises cannot secure AI agents by making the underlying models more robust and should instead enforce security controls at the system level. The authors distilled five principles from systems security (least privilege, tamper resistance of the trusted computing base, complete mediation, secure information flow, and the human as a weak link) and analyzed eleven real-world attacks on AI agents, all of which violated secure information flow.

OpenAI announced it was named a Leader in the Gartner Magic Quadrant for Enterprise AI Coding Agents, citing Codex's broad developer surface and enterprise controls such as approval gates, RBAC and OS-level sandboxing. The company says Codex is used by more than 4 million people each week, including customers such as Cisco, Datadog, Dell Technologies and NVIDIA. The Gartner report is dated April 2026.