antoan.ai (AI Safety Vietnam)

antoan.ai (also written AnToàn.AI — an toàn meaning "safety" in Vietnamese) is Vietnam's first AI-safety community organisation, dedicated to building the local pipeline for AI-safety research and practice. Its mission is to "support and create opportunities for scholars, students, and professionals in Vietnam to exchange and practise in the field of AI safety, from technical to policy perspectives through translation projects, hackathon/reading group events, and research practice." The group operates with dual in-person hubs in Hanoi and Ho Chi Minh City and a distributed online presence via Discord and newsletter. It is the corpus's first Vietnam entry and the first local-group entry for mainland Southeast Asia.

Team

antoan.ai is a nonprofit with a small team. Identified members include Parker Vu, Tung Nguyen Lam, and Khiem Hoang (a computer science student at Waseda University); a broader collaboration network named in the Summer Research Fellowship includes Jord Nguyen, Khanh Linh Nguyen, and N Tran. The founding date is not publicly documented; the Summer Research Fellowship run in partnership with the Hanoi AI Safety Network was underway by at least the early 2020s.

Programme

antoan.ai's work runs across four areas.

Translated resources. The centrepiece is the AI Safety Handbook (Cẩm nang an toàn AI), a Vietnamese-language translation of the handbook compiled by researchers at the French AI Safety Center (CeSIA). Alongside it: an AGI Career Guide (Cẩm nang nghề nghiệp AGI, drawing on 80,000 Hours and AI Safety Atlas frameworks), an AI-safety terminology glossary, and a regular newsletter — all in Vietnamese, addressing a significant language-access barrier to global AI-safety materials.

Summer Research Fellowship. A technical AI-safety fellowship run in partnership with the Hanoi AI Safety Network, covering safety evaluations, interpretability, AI control, and adversarial robustness. Previous participants have published research at ICLR, ICML, and AAAI — a concrete pipeline from community participation to credentialed technical output.

Global South AI Safety Hackathon. For the 2026 edition of the Apart Research Global South AI Safety Hackathon (June 20–21, 2026), antoan.ai serves as the organising body for Vietnam's two jam sites — Hanoi and Ho Chi Minh City — making it the sole Southeast Asian hub among eight global locations (others span Latin America, Cape Town, and Bengaluru). The Asia track prize pool is USD 2,000; hubs provide co-working space, refreshments, and streamed keynote and mentorship sessions.

Community layer. Reading groups, a Discord community, and events connecting participants across technical and policy aspects of AI safety.

Place in the movement and civic context

antoan.ai closes the mainland Southeast Asia gap in the corpus's local-group layer — Singapore AI Safety Hub (SASH) covers a city-state; antoan.ai covers Vietnam's continental footprint. It sits in the same EA-adjacent insider-community pattern as AI Safety India, AI Safety UAE, Berlin AI Safety, and AI Safety 東京: groups whose central activity is building the local practitioner pipeline for AI-safety work. What distinguishes antoan.ai is its translation-first approach — the resource library in Vietnamese serves a population where English-language AI-safety materials present a real barrier — and its dual-city hub structure, reflecting Vietnam's split cultural geography between Hanoi (the political capital) and Ho Chi Minh City (the commercial hub).

The civic context warrants honest acknowledgement. Vietnam is a one-party state in which civil society operates under significant constraints: Decree 126 (2024) tightened rules on association formation and limited foreign funding, and Global Voices reported in January 2025 that nearly a dozen NGO workers had been detained in the preceding four years. The frame that applies to antoan.ai is comparable to the one noted for AI Safety UAE in a different single-party-adjacent context: AI-safety community work is not in tension with state priorities — it is substantially aligned with them. Vietnam enacted its first binding AI law (Law No. 134/2025/QH15) on 10 December 2025, effective 1 March 2026, establishing a risk-based framework with human-oversight requirements; building domestic AI-safety research capacity and governance literacy is the kind of work the law's implementation will benefit from. antoan.ai's technical focus (alignment, interpretability, AI control) and its international partnerships (Apart Research, CeSIA, Hanoi AI Safety Network) situate it within the global civil-society AI-safety movement while affording it operational space that more directly political or rights-focused groups in Vietnam would not enjoy.