Agent Infrastructure: Data, Voice, Payments, OS, and Sandbox Risk

Now everyone can put data to work. OpenAI adds a Data agent to ChatGPT Work, letting employees query company data, build interactive dashboards, and take action without SQL. Outcome engineers get an enterprise-grade semantic layer and a direct agentic path from natural-language intent to grounded actions — useful when you need reproducible data-to-action artifacts and operable knowledge graphs (Principles 03, 11, 06).

Build more natural voice experiences with GPT‑Live‑1 in the API. OpenAI releases GPT‑Live‑1 enabling full‑duplex voice agents that handle interruptions, cut latency, and delegate complex reasoning to backend models. This changes agent UX and orchestration: voice becomes a low-latency channel that demands new session lifecycle, fallback, and validation patterns in your agent stack (Principles 09, 07).

Microsoft 2.5: EVP Pavan Davuluri wants to remake Windows for both human and agent users. Microsoft designs Windows with new identity, security primitives and Execution Containers to natively run agentic workloads alongside humans. Outcome engineers should plan for OS-level primitives that change deployment, identity, and isolation models — treat the desktop/host as part of your agent platform design and threat model (Principle 07).

Ant International, Visa, and Mastercard plan a new standard for payments via AI agents. Ant, Visa, and Mastercard announce a joint standard to let AI agents transact payments securely at scale. If your agents will take actions with economic consequences, design for credentialed agent wallets, immutable audit trails, and human review gates to meet interoperability and compliance expectations (Principles 09, 15).

Claude Code, Codex, and Cursor Have Leaky Sandbox Problems You Don’t Hear About. Accomplish warns of widespread sandbox‑escape vulnerabilities in Claude Code, Codex, and Cursor and slow fixes from vendors. Outcome engineers must assume execution sandboxes are brittle: add defense‑in‑depth, runtime monitoring, and enforceable review gates in your software factory to prevent rogue agents and data exfiltration (Principles 14, 15, 16).