Melbourne Airport deploys AI agents for incident response — and finds governance gaps

Why it matters: A real-world operations rollout shows the upside of agentic automation (faster SOP retrieval, automated shift reports) and the immediate need for access controls to prevent sensitive data exposure — exactly the balance most enterprises must strike as they scale AI Employees.

Melbourne Airport’s data team integrated autonomous agents with SharePoint to speed incident response across runways and terminals and to auto-assemble executive shift reports. Early wins: faster decision support and less manual reporting. Early lessons: agents surfaced sensitive personal information from OneDrive, prompting a permissions review and tighter governance. Takeaway for operators: before unleashing agents on enterprise content, harden identity and access, set retrieval boundaries, and log/attest agent actions for auditability.

Gen Digital launches VPN for Agents and expands Norton AI Agent Protection

So what: Consumer and SMB agent usage is rising, but network trust is lagging. A purpose-built “VPN for Agents” with real-time oversight can reduce data exfiltration and policy drift as autonomous tools browse, fetch, and post on your behalf.

Gen (Norton, Avast, LifeLock) added an AI-native VPN that authenticates and sandboxes agent traffic, alongside Norton 360’s AI Agent Protection for runtime monitoring. Expect similar “trust layers” to become table stakes as agents gain persistent permissions and multi-app control.

TestMu AI debuts Kane CLI to test code produced by AI agents

So what: As agents write more front-end and workflow code, verification must be automated. Command-line browser runs let CI pipelines “trust but verify” before agent-written code hits production.

Kane CLI executes browser flows from the terminal to validate agent-generated code and close the gap between code suggestion and working features. For SMBs adopting AI dev copilots or agents, pair generation with deterministic tests to prevent silent breaks.

EU’s €20B AI compute “gigafactories” face backlash over demand and dependency

Why it matters: If Europe misallocates compute funding, startups could still rely on U.S. clouds while taxpayers foot the bill. For companies operating in the EU, watch for where capacity lands and whether access terms actually favor SMEs.

Politico reports mounting criticism of the Commission’s plan to fund four to five mega facilities (100,000+ GPUs each). Skeptics question real demand beyond a handful of model labs and warn about deepening Nvidia dependence. The Commission frames the push as sovereign compute for European industry and research. Procurement and access design will determine whether this becomes a catalyst or a costly white elephant.

Chinese court: Cutting jobs because “AI is cheaper” is illegal

Why it matters: Expect more legal scrutiny of AI-driven restructuring. Even if productivity improves, employers may need reskilling or redeployment plans — not blanket terminations justified by automation.

The Hangzhou Intermediate People’s Court ordered compensation of over 260,000 yuan to a worker dismissed on AI cost grounds, echoing a similar 2024 ruling. The court said AI has not “substantially replaced” human work. Signal for multinationals: document role redesign and provide fair process; “AI as pretext” is increasingly risky beyond China as regulators and courts catch up.

Pixxel x Sarvam to launch AI-powered orbital data centre satellite

Why it matters: Edge AI is heading to space. Onboard GPUs + hyperspectral imaging could deliver real-time insights for agriculture, mining, and climate — with lower downlink costs and faster decisions.

Indian space-tech firm Pixxel will build and operate “Pathfinder,” a ~200 kg satellite carrying datacenter-grade GPUs to run Sarvam’s language models in orbit. By analyzing hyperspectral data onboard, operators can prioritize only high-value results to Earth — a preview of space-based compute for time-sensitive, data-heavy workloads.