Singapore AI Safety Hub (SASH)

The Singapore AI Safety Hub (SASH) is a Singapore-anchored AI-safety coworking space, events platform, and ecosystem-builder whose mission is to develop a civil-society AI-safety field in Asia. Founded in early 2025 by Miro Plueckebaum — a former Product and Governance Lead at Conjecture and former Product Lead at ByteDance in Beijing and Singapore, who also works as Strategy & Research Manager at the Oxford Martin AI Governance Initiative — SASH describes itself as "an AI safety ecosystem builder and research organisation" that bridges Eastern and Western AI-governance perspectives. Its central premise is that Singapore, as an AI governance hub in Southeast Asia and a frequent setting for international AI-safety convenings, is a natural base from which to build the regional practitioner community that AI-safety work in the Asia-Pacific requires. It is the corpus's first entry covering a civil-society AI-safety community group in Southeast Asia, and is rated "Very Active" by the global AISafety.com community directory.

Founding and structure

SASH was established in 2025 and is structured as a nonprofit. Its founder, Miro Plueckebaum, splits his time between managing partnerships at SASH and his work at the Oxford Martin AI Governance Initiative and the Centre for the Governance of AI (GovAI). The operational team listed publicly includes Sam Boger, Raymund Ed Dominic Bermejo, Valerie Pang, Oskar Galeev, and Zachary Richardson, giving SASH a notably larger staff footprint than most community groups at this stage — closer to a small organisation than an informal meetup, a shape that reflects both the ambition of building a regional hub and the demands of running a physical co-working space.

The hub's primary physical base is WeWork at 22 Cross Street, Singapore 048421, near Chinatown MRT — a central location that positions SASH within reach of Singapore's research university campuses, government tech agencies, and international corporate AI labs. Community communications run via WhatsApp alongside the Luma events calendar. The hub solicits contact from interested researchers at hello@aisafety.sg.

Three-pillar programme

SASH organises its work under three named pillars: field development in Asia (co-working, events, skill-building, and policymaker engagement), global collaboration (convenings, fellowships, and research projects connecting Singapore with international AI-safety organisations), and specialised programmes (intensive training and technical research initiatives). The three-pillar structure reflects the hub's dual mandate: building a local practitioner community while serving as a convening point for international AI-safety activity arriving in Singapore.

Events — research exchanges, hackathons, and convenings

The hub's public launch featured a lightning-talk event at 22 Cross Street at which five Singapore-based practitioners spoke: James Chua (Truthful AI) on AI that does not lie, Siao Si Looi (FAR.AI) on safety research that bridges academia and commercialisation, Clement Neo (Apart Research) on mechanistic interpretability and research incubation, Robin Staes-Polet (The Future Society) on AI governance, and Nicholas Chen (PlethorAI) on agent security — a programme that signalled the range of technical-safety, governance, and security work SASH intends to anchor in Singapore.

The hub subsequently ran its inaugural AI Safety Research Exchange, a practitioner-facing session featuring six presentations covering confidence elicitation as an LLM attack vector, mechanistic interpretability in language model refusal systems, reasoning capabilities in large language models, persona-based vulnerabilities, multi-modal model interpretation, and modular approaches to safe AI development. The exchange welcomed interested researchers via the hub's contact email, establishing it as a recurring programme format.

SASH has served as a recurring local jam site for Apart Research hackathons. The AI Control Hackathon 2025 (March 28–30, 2025), co-organised by Apart Research and co-hosted by IMDA and Concordia AI, ran three research tracks at WeWork 22 Cross Street: ControlArena Challenges (using the UK AI Safety Institute's framework), Control Protocol Design, and Red Teaming and Vulnerability Research, with $2,000 in prizes. SASH also hosted an in-person jam site for the AI Safety × Physics Grand Challenge (July 25–27, 2025) and is listed as a host for the AI Control Hackathon 2026, indicating a sustained partnership with Apart Research's global sprint programme.

Beyond its own events, SASH co-hosted the AI Safety Social on January 23, 2026 at 32 Carpenter Street — an informal networking gathering held as a side event of Singapore AI Research Week coinciding with AAAI 2025, co-organised with IMDA and Concordia AI, and described as a forum connecting researchers, practitioners, and policymakers from AI-safety organisations globally.

AI Security Bootcamp

One of SASH's most resource-intensive programmes is its institutional support role for the AI Security Bootcamp — a 7-day intensive training programme targeted at experienced security professionals, covering threat modelling for advanced AI systems, adversarial attacks, LLM-specific risks (jailbreaks, prompt injection), hardware governance, and AI control mechanisms. The inaugural Singapore edition, scheduled for April 20–26, 2026, was led by Pranav Gade (Conjecture) and Nitzan Shulman (Heron AI Security Initiative), with SASH providing local execution and institutional support. The programme deliberately overlapped with Black Hat Asia (April 21–24) and ran immediately before DEF CON Singapore (April 28–30), situating AI-safety security training within Singapore's broader international cybersecurity calendar.

Place in the movement

Singapore AI Safety Hub is the corpus's first Southeast Asia entry for a civil-society AI-safety community group — the region's equivalent of the insider-community groups the corpus covers in Berlin, Tokyo, and Cape Town. What distinguishes SASH from those counterparts is scale of ambition and organisational form: where Berlin AI Safety and AI Safety 東京 began as volunteer-run monthly meetup groups and grew into richer programme platforms, SASH launched as a structured nonprofit with a salaried team, a physical co-working facility, and an explicit Asia-hub positioning — the aspiration is to serve as the connective tissue between Singapore's government AI-safety infrastructure (the Singapore AI Safety Institute, the SCAI conference, IMDA's safety programmes) and the international civil-society AI-safety practitioner community. This reflects a structural feature of Singapore's position in the regional AI landscape: unlike Tokyo or Berlin, Singapore hosts frequent international AI-governance convenings (SCAI 2025 ran alongside ICLR 2025), making it a natural relay point for institutions like GovAI and the Oxford Martin AIGI that are already present in Plueckebaum's portfolio.

The insider-community function SASH performs — low-barrier research-exchange events, co-working access, hackathon hosting — parallels the community-building substrate that AI Safety 東京 provides in Japan and Berlin AI Safety provides in Germany. The security-bootcamp and technical-hackathon programming, however, brings in a defensive-security and AI-control research strand that the corpus's other insider-community groups do not yet emphasise — a strand that maps to the practitioner backgrounds (Conjecture, FAR.AI, Heron AI Security) of the Singapore network SASH has assembled around itself.