From Whitepaper to Weather Emergency: OpenAI's Bangkok AI Jam Goes Full Survival Mode

In Bangkok today, OpenAI convened 50 disaster management leaders from 13 countries across Southeast and South Asia for its first AI Jam focused on disaster response. The event brought together government agencies, multilateral organizations, and non-profits—including participants from Bangladesh, India, Indonesia, Lao PDR, Malaysia, Myanmar, Nepal, Pakistan, Philippines, Sri Lanka, Thailand, Timor Leste, and Vietnam. Held with support from the Gates Foundation, the Asian Disaster Preparedness Center, and DataKind, the gathering tackled a pressing question: can AI actually help institutions respond faster and more effectively when disasters strike? Basically, can ChatGPT save lives before it finishes generating your horoscope?

The initiative builds on the OpenAI for Countries Program announced at Davos, aiming to move past experimentation and embed AI into daily operational challenges across Asia. Disaster response teams often operate under severe constraints—fragmented data, manual processes, and limited infrastructure that slow coordination during rapidly evolving crises where timely information is critical. We're not talking about optimizing ad spend here; we're talking about situations where delays cost lives and spreadsheets are basically medieval technology holding back people who need to move at machine speed.

Many organizations at the Jam are already testing humanitarian AI solutions, with careful attention to integrating new tools alongside existing systems rather than replacing local expertise. The vibe wasn't "AI is going to save us all" but rather "let's not throw out the baby with the bathwater while the building is on fire." Smart approach—nothing says "we learned nothing from the Web3 pivot bro era" like showing up to a disaster zone with a solution looking for a problem.

The urgency is real. Asia remains the world's most disaster-prone region, accounting for approximately 75% of people affected by disasters globally. The World Bank estimates disasters have cost ASEAN countries more than $11 billion in previous years. In the second half of last year, a series of typhoons and severe storms tested response systems across South and Southeast Asia, exposing gaps in data, coordination, and surge capacity. That's not a problem statement—that's a market opportunity, if you're the kind of person who thinks about natural disasters in terms of TAM.

Meanwhile, people are already turning to AI during emergencies. During Cyclone Ditwah in Sri Lanka, cyclone-related messages on ChatGPT increased 17×. In November 2025, during Cyclone Senyar in Thailand, message volume rose 3.2× compared with preceding months. These spikes suggest an opportunity to align AI tools more closely with official disaster communication and field operations. Turns out when the sky turns gray and the wind starts howling, humans do what they always do: ask machines for help. The question is whether the machines are actually ready for prime time.

At the Bangkok Jam, participants worked with OpenAI mentors to develop practical applications—from custom GPTs for responders to reusable workflows across different scenarios. Teams explored tools for situation reporting, needs assessments, and public communication: processes that often become bottlenecks during major disasters. Sessions also emphasized responsible use, transparency, and safeguards, with a focus on building institutional trust so staff feel confident using new systems in high-stakes environments. Nobody wants to be the responder who trusted an AI that confidently hallucinated evacuation routes during a Category 4 storm.

Sandy Kunvatanagarn, Head of Public Policy at OpenAI, framed