Summary

  • Amajit Gupta’s APNIC 62 keynote proposed a network “autonomous grid” built around real-time telemetry, programmable paths and self-correcting systems, while keeping human-defined guardrails and oversight.
  • Yukako Uchida’s JPCERT/CC presentation described another operating requirement: incident responders need trust built through repeated contact before organizations can coordinate safely under pressure.

The first response to a deteriorating network path may need to happen in milliseconds. The first dependable conversation between two incident-response teams may have begun years earlier. APNIC 62 placed these clocks in the same keynote session, without turning them into one problem or proposing a shared operating rule.

Amajit Gupta, Lightstorm’s Group CEO and Managing Director, framed latency as the scarce resource for workloads spread across clouds, data centres and edge systems. His 8 September presentation, “The Autonomous Grid,” set out four capabilities: real-time telemetry, intent-based orchestration, programmable path control, and resilient self-correcting architectures. He argued that network operations need to respond dynamically rather than wait for manual intervention. The deck also acknowledges the harder work: adapting legacy infrastructure, operating processes and staff skills.

Gupta did not describe a measured deployment or a verified performance gain. His presentation is a proposal for how networks might operate. It also does not make human judgment obsolete. He argued that closed-loop automation should work within human-defined guardrails and leave engineers responsible for architecture, resilience, governance and oversight. That boundary matters: a system that detects a condition and changes a path is not automatically authorized to alter another organization’s service or share sensitive information with it.

Yukako Uchida, JPCERT/CC’s Manager of Global Coordination Division, spoke about the human work behind cybersecurity collaboration. Her presentation says JPCERT/CC handled approximately 20,000 incidents in 2025. It also records incident-report contacts across 102 economies and notifications to 167 economies over 20 years, a measure the slide limits to ccTLDs interacted with for incident response. The figures describe the scale of coordination, not a trend in attacks or a measure of successful outcomes.

Uchida’s point was not that formal structures or technical tools are useless. Her slides emphasize that responders need confidence that counterparts will handle shared information responsibly and act appropriately. That confidence develops through first-hand, repeated interaction. Community groups and exercises can create opportunities for that contact, but attendance itself cannot prove that a participant has authority to make a binding decision for an organization.

This is where the two keynotes meet. Telemetry can show that a path is degraded; orchestration can select an alternative. An incident report may still require a human to judge what can be shared, whether the sender is trusted, and which organization can authorize containment. These are different control surfaces. The first sits within an operator’s configured network; the second crosses institutional boundaries. Faster automation can shorten the time available for the second decision, but it cannot supply the missing authority or relationship by itself.

The conference recap presents both talks as part of a future that needs technological innovation and stronger communities. That is a useful juxtaposition, not evidence of a joint proposal, pilot or deployed “autonomous grid.” The practical test is whether operators can explain what an automated system may do, what it must escalate, how an action is logged, and how a mistaken cross-network change can be reversed. APNIC’s published material does not establish that those controls are already in place across operators.

The stakes are therefore specific. A network may be able to react faster than an engineer can read an alert, while an incident-response team may need to establish trust before it can act on another party’s report. Treating those as the same kind of automation would blur who owns the decision. Treating them as separate but connected processes makes the handoff visible: machine speed inside a control boundary, accountable human authorization where one organization’s action reaches another.

Sources: APNIC’s APNIC 62 keynote recap; Amajit Gupta’s Autonomous Grid presentation; Yukako Uchida’s cybersecurity collaboration presentation.

APNIC 62 Conference Report.