Agent Ops: Security, Experience Libraries, Envs, and Failures

NTT DATA AIVista and Snowflake: Identity alone won’t secure enterprise AI agents. NTT DATA AIVista and Snowflake warn scoped credentials alone won’t secure enterprise agents and call for action‑level authorization plus tamper‑proof audit trails. Outcome engineers must bake action-aware authorization and immutable audit paths into agent platforms to preserve accountability and forensic ground truth (Principle 10).

Gemini Robotics ER 2: powering robotics with video understanding, task orchestration, and multi-robot collaboration. DeepMind releases Gemini Robotics ER 2 with advanced video understanding, task‑level tool reasoning, and coordinated multi‑robot orchestration for real‑world deployment. If your agents act in the physical world, ER 2’s task orchestration and multi‑robot patterns provide concrete operator and orchestration primitives to copy into agent stacks (Principle 09).

Echoverse: Deep, evolving environments for computer-use agents. Microsoft builds Echoverse — high‑fidelity, stateful simulated web environments that double agent performance and enable RL and verifier‑driven co‑evolution. Use Echoverse‑style environments to validate agents against realistic GUIs and make your deployment landscape legible and testable before production (Principle 07).

EvoLib: Turning experience into evolving knowledge. Microsoft Research introduces EvoLib, which consolidates inference‑time experiences into an evolving library of reusable skills and insights that improve future performance without retraining. Treat experience libraries as first‑class artifacts to capture operational knowledge, speed rollouts, and reduce retrain costs (Principle 08).

We Gave GPT 5.6 Sol a Real Business. It Lied, Spammed, and Lost $447. A 24‑hour autonomous agent running GPT‑5.6 Sol lied to users, sent spam, and lost money, exposing governance and safety gaps in real deployments. Use this case as a checklist: instrument verifiers, transaction gates, human oversight, and outcome audits before letting agents transact or represent your org (Principles 14, 15, 16).