OpenAI Pushes Disaster-Response AI From Demo Culture Toward Field Use in Asia

AI disaster response is moving closer to operational reality

Artificial intelligence has been discussed for years as a tool that could help in emergencies, but much of that discussion has remained abstract. OpenAI is now trying to shift the conversation toward practice. In Bangkok, the company is holding what it describes as its first AI Jam for disaster management professionals, bringing together 50 leaders from across Southeast and South Asia.

The workshop is being run in partnership with the Gates Foundation, the Asian Disaster Preparedness Center, and DataKind. Participants come from 13 countries: Bangladesh, India, Indonesia, Lao PDR, Malaysia, Myanmar, Nepal, Pakistan, Philippines, Sri Lanka, Thailand, Timor Leste, and Vietnam. According to OpenAI, they represent government agencies, multilateral organizations, and nonprofits directly involved in disaster response.

That mix is important. It suggests the event is not designed as a general AI awareness exercise. It is aimed at organizations that already work under time pressure, in resource-constrained environments, and with fragmented information flows. Those are precisely the conditions in which the value of AI must be judged by usefulness rather than novelty.

Why Asia is the focal point

OpenAI’s post makes a straightforward case for the regional focus. Asia, it says, remains the world’s most disaster-prone region, accounting for an estimated 75% of people affected by disasters globally. It also cites a World Bank estimate that disasters have cost ASEAN countries more than $11 billion in previous years.

Those figures help explain why disaster management is becoming an increasingly plausible entry point for practical AI. The need is recurrent, the stakes are high, and many response teams work with limited infrastructure and manual processes. According to OpenAI, those constraints can slow coordination and delay decisions, especially when timely information matters most.

That description fits a large part of the real emergency-response challenge. In many crises, the core problem is not a lack of data in the abstract. It is the difficulty of turning fragmented, fast-moving information into something response teams can actually use.

Evidence of public demand during emergencies

OpenAI also points to a signal that is easy to miss but potentially important: people are already turning to AI during disasters. The company says internal data showed a 17-fold increase in cyclone-related messages on ChatGPT during Cyclone Ditwah in Sri Lanka. During Cyclone Senyar in Thailand in November 2025, message volume related to the event increased 3.2 times compared with prior months.

Those numbers do not prove that AI is already integrated effectively into official disaster systems. But they do show that demand for AI-based information support can spike rapidly when crises hit. That has two implications. First, people appear willing to use these tools under stress. Second, emergency-response institutions may increasingly need to think about how public-facing AI use intersects with official workflows.

In other words, the question is no longer simply whether AI should play a role in emergencies. In some cases, it already is. The more urgent question is whether the role will be improvised by users or intentionally integrated by professionals.

From interest to embedded use

OpenAI frames the workshop as part of a broader effort to help organizations move beyond interest in AI and toward real-world applications. That is a useful distinction. The hardest phase of adoption is rarely awareness. It is translation: turning a tool with broad capabilities into something dependable inside a specific mission.

Disaster response is full of that translation challenge. Teams need to gather information, coordinate people, support affected communities, and make time-critical decisions. AI can sound promising in each of those areas, but promise alone is not an operational plan. The gap between a capable model and a usable emergency workflow is where most implementation efforts either mature or stall.

By convening people already responsible for response, OpenAI and its partners appear to be testing a more grounded path. Rather than ask what AI can do in theory, they are asking what it can do inside the constraints responders already face every day.

A shift away from AI theater

The deeper significance of the Bangkok workshop is that it represents a move away from AI theater and toward narrower, accountable use cases. Disaster management is not a forgiving domain. Tools either help teams move faster and think more clearly, or they create noise. That forces a higher standard.

It also aligns with the company’s broader OpenAI for Countries effort, which it says was expanded at Davos. The implication is that national and regional use cases, rather than purely consumer novelty, are becoming a more visible part of the company’s public agenda.

Whether these efforts produce durable deployments remains to be seen. But the workshop makes one thing clear: in a region repeatedly hit by storms and disasters, AI’s next test will not be whether it can impress in a demo. It will be whether it can help response teams do better work when time and information are both in short supply.

This article is based on reporting by OpenAI. Read the original article.