SINGAPORE, 15 October 2026 – Vulcan announced that it has been accredited under the AI Tester Accreditation Programme (AI TAP) run by AI Verify Foundation (AIVF). The announcement was made at AI Assurance Exchange: The Agentic AI Edition, hosted by AIVF in Singapore.
AI TAP recognizes testing service firms that can technically assess AI systems for safety, reliability and governance risks. AIVF describes it as the first program of its kind in Asia. Firms are assessed on technical competency, validated against guidelines such as IMDA’s Starter Kit for Testing LLM-Based Applications for Safety and Reliability, as well as on good standing, financial sustainability, operational readiness and whether their declared business scope matches actual capabilities.
A new direction for AI in Singapore’s financial sector
The accreditation announcement comes days after the Monetary Authority of Singapore (MAS) issued its new Guidelines on Artificial Intelligence Risk Management, which apply to all financial institutions (FIs) and all forms of AI, including generative AI and AI agents.
The Guidelines place testing at the center of AI risk management. MAS expects FIs to run vulnerability assessments, penetration testing and red teaming across their AI systems, and to test generative AI and agents against their key failure modes, including data leakage and adversarial attacks. Where FIs rely on AI from third-party providers, MAS accepts independent certifications or external assessments by competent parties, but not vendor self-attestations.
The need for Independent, competent testers is exactly the gap AI TAP is designed to fill. For FIs, working with a tester recognized under AI TAP offers an evidence-based way to demonstrate that their AI testing is rigorous.
How Vulcan aligns with the MAS Guidelines
Vulcan works with major commercial banks and financial holding companies across the region and designs its testing around the expectations set out in the Guidelines. Its offerings map to several of the testing and control areas the Guidelines describe:
- Adversarial testing and red teaming (Sections 5.14(b), 5.15, 5.22(b)). Vulcan Attack automates AI red-teaming for generative AI applications and agents, testing both single model responses and multi-step agent workflows against manipulation, evasion and data leakage, and checking whether guardrails hold under attack. Testing also covers multiple languages, with native-speaking testers working in English, Chinese, Arabic, Hindi, Japanese, Korean, Thai, Indonesian Bahasa, Malay Bahasa, Vietnamese and other regional languages.
- Securing AI inputs and outputs (Sections 5.4(e), 5.16). Vulcan Protect provides real-time runtime protection, applying input validation and output controls to reduce data exposure.
- Agent oversight and control (Section 5.23). Vulcan Harness, Vulcan’s agentic enforcement layer, applies policy controls to AI agents at runtime, helping FIs oversee the actions agents take and the tools they use.
- Independent third-party assurance (Sections 5.10, 5.11). As an AI TAP-accredited tester, Vulcan can support FIs that need external assessments of vendor AI, as well as AI vendors that need independent evidence when selling to FIs.
Alvin Kwock, Co-Founder and CEO of Vulcan, said, “MAS has made it clear that financial institutions need independent, rigorous testing of the AI they build and buy, especially as agents take on more autonomous work. Being accredited under AI Verify Foundation’s AI Tester Accreditation Programme reflects our commitment to testing that is transparent, evidence-based and grounded in real-world attack techniques. As organizations plan for 2027, we hope to see AI testing, guardrails and harness built into AI project planning from the start. We look forward to helping financial institutions and AI adopters in various sectors in Singapore and across the region prepare for the new Guidelines.”
