Trustworthy AI Investment as Governance

Contents

Executive Summary

This panel discussion argues that capital allocation is a critical governance tool for AI development, with investment flows determining which AI systems get built, who benefits, and what safety standards are prioritized. The panelists highlight a severe global asymmetry: billions flow into frontier model development while funding for safety infrastructure, standards, testing, and AI adoption in the Global South remains critically underfunded. The core message is that governments, investors, and boards must actively reshape market incentives through procurement, regulation, certification, and impact investment to align capital with trustworthy AI principles.

Key Takeaways

Capital allocation is governance. Where money flows determines what gets built and prioritized. Without intentional capital reallocation, regulations and safety frameworks remain unenforced and marginalized.
Use existing government tools immediately. Procurement requirements, tax incentives, mandatory disclosure standards, and public funding already work in other sectors (climate finance, aviation safety). They need rapid adoption for AI.
Certification and standards unlock private investment. Third-party assurance mechanisms (like Rosaro in Singapore) give investors confidence and reduce friction for corporate adoption—creating self-reinforcing market incentives for trustworthy AI.
Human capital is the bottleneck, not just money. Funding billions means little without trained governance professionals, AI safety researchers, and informed board members. Skills development must be co-invested alongside infrastructure.
Hope and opportunity must accompany governance discussions. Without framing AI governance as enabling beneficial outcomes rather than restricting innovation, adoption will lag and public distrust will grow. Optimism about responsible AI is essential to market-wide change.

Key Topics Covered

Capital allocation as AI governance — Investment flows as more powerful than regulations in shaping AI development
Funding asymmetries — Massive disparity between frontier model investment vs. safety, standards, and Global South innovation
Market incentives and regulatory tools — Government procurement, tax breaks, subsidies, and mandatory disclosure requirements
Certification and standards frameworks — Third-party verification, accreditation, and compliance mechanisms as market drivers
Human capital and skills development — Critical gaps in workforce training, particularly in developing regions
Board accountability and legal responsibility — Personal liability models and board-level governance structures
Consumer demand and bottom-up pressure — Role of public awareness and stakeholder demands in driving change
Regional disparities — Differences between Singapore's approach and broader Global North/South investment gaps
Safety-by-design and trustworthiness by default — Integration of safety into product development rather than post-hoc compliance
Impact venture capital — Double bottom-line investing that balances commercial returns with societal benefit

Key Points & Insights

Capital determines direction, not regulation. As Muhammad Nab (Modilla Ventures) emphasizes, "it's really capital that's deciding what's trustworthy and what's investable." Regulations lag significantly behind market-driven investment decisions, and competitive pressure among frontier labs actively discourages investment in safety.
Massive funding asymmetry exists across AI investment types. Billions flow to frontier model development (OpenAI, Anthropic, etc.) while safety infrastructure, open-source tooling, testing companies, and AI adoption in developing economies receive a tiny fraction. The panelists stress this is not accidental but reflects current market incentives.
Safety investment is growing but insufficient. There is a "growing trend on AI safety investment" but it remains far below proportional allocation compared to model training and infrastructure—analogous to how airlines invest equally in safety as in other engineering domains.
Government procurement is an underutilized lever. Singapore's accreditation program demonstrates how government buying power can de-risk and accelerate adoption: companies with government certification gain faster procurement access and reduced friction, creating strong market incentives.
Certification mechanisms are critical but underfunded. The absence of standardized, trustworthy AI certification comparable to financial auditing creates friction for corporate adoption. Board members and executives lack assurance mechanisms and currently rely on ad-hoc due diligence.
Global South lacks foundational capacity. Funding gaps aren't just about safety—they affect AI adoption in education, healthcare, and financial services in developing economies. Combined with skills shortages and connectivity gaps, this perpetuates innovation inequality.
Market signals from major investors reshape behavior rapidly. Gabriella Ramos cites BlackRock's 30% board gender requirement as driving structural change in corporate governance faster than decades of prior advocacy—illustrating investor power to reshape norms.
Personal accountability of executives accelerates change. Spain's legal responsibility framework for social network platform owners creates board-level pressure for compliance faster than regulation alone. Personal legal risk concentrates minds.
Trust cannot be "tickboxed" in PDFs—it must be embedded in design. As Krishna Gad (quoted by Muhammad) noted, trustworthiness must be built into every runtime system layer (edge, on-premises, cloud), not added post-deployment through compliance documents.
Investment must span the entire capital stack. Solutions require funding downstream (startups, open-source), midstream (testing, certification, standards bodies), and upstream (sovereign wealth funds, institutional investors) to shift incentives at every level.

Notable Quotes or Statements

"Money is not going into trust. So this is why this is a really exciting conversation." — Muhammad Nab, emphasizing the market gap that creates opportunity for impact investors.
"Governance is slow and we talk about closed doors about all those principles, but governance is in the capital flows." — Amir (moderator), reframing governance as a market mechanism rather than bureaucratic process.
"Trust can't be left in a PDF—it needs to be embedded in the runtime." — Krishna Gad (cited by Muhammad Nab), capturing why post-hoc compliance fails and why safety-by-design is essential.
"When I had my airplane ticket, nobody told me there's a 5% chance it will fall. Would I have taken the plane?" — Amir, using aviation safety analogy to illustrate why AI governance requires proportional investment.
"If your employees are afraid they're going to lose their job from the exact thing you got them working on, I don't think you're going to realize those returns." — Alpesha, highlighting how fear undermines organizational adoption of AI governance.
"Capital flows will decide which AI system is built, who will benefit from AI systems, how safety will be developed, how inclusion will be developed." — Sophie (opening framing), summarizing the panel's core premise.
"We know mechanisms exist. We know incentives exist in the general economy. The issue is why are we not using them?" — Gabriella Ramos, on the availability of proven policy tools waiting for application.
"Successful companies will be able to put a label close to their name saying 'I'm AI ready' or 'I'm AI certified.'" — Julian Bio, envisioning a market where certification becomes a competitive advantage.
"We should be walking away with hope, opportunity—not fear." — Alpesha (closing), reframing the tone of AI governance discussion.

Speakers & Organizations Mentioned

Speaker	Organization/Title	Role
Muhammad Nab	Modilla Ventures	Managing Partner (VC investing in responsible AI)
Sophie	Global Partnership on AI	Panel co-host; discussion facilitator
Amir	(Not fully identified)	Panel moderator
Alpesha	IEEE/IMDA (standards)	Head of AI standards
Julian Bio	SCAI (Canada)	CEO (nonprofit supporting AI adoption & trust)
Gabriella Ramos	UNESCO; G7 Task Force on Inequalities	Co-chair; former Assistant Director-General UNESCO
Speaker from Singapore	IMDA (Infocomm Media Development Authority)	Policy/governance official for Singapore
Krishna Gad	(Company unspecified)	AI governance entrepreneur (quoted)

Institutions Referenced

Global Partnership on AI — Publishing investment data on AI funding
Modilla Ventures — Impact VC fund in responsible AI
IEEE/Standards bodies — Developing certification frameworks (e.g., IEEE 7000 series)
SCAI — Canadian nonprofit supporting trustworthy AI adoption
IMDA Singapore — Government agency running accreditation programs (Rosaro subsidiary; Laurang AI space)
Tamasic — Singapore sovereign fund
UNESCO — Policy development on AI inequality and disclosure
G7 — Policy coordination on AI safety and financial disclosure

Technical Concepts & Resources

Standards & Frameworks

IEEE 7000 series — Ethical AI standards (mentioned by Alpesha)
IEEE Ethically Aligned Design — Certification framework used by governments and companies
AI Ethics Governance Certification Program — IMDA/IEEE tool enabling capacity building and corporate value demonstration

Tools & Mechanisms

Third-party assurance providers (e.g., Rosaro in Singapore) — Independent AI safety and trustworthiness testing companies
Accreditation programs — IMDA's model for vetting startup products on security, reliability, and sustainability
Laurang AI — Singapore's physical innovation space for startup collaboration and learning

Organizational Structures

Impact venture capital (double bottom-line investing) — Balancing commercial returns with measurable societal impact
Board certification and training — Programs to educate board members on AI governance (mentioned in Singapore context)
Task forces on financial disclosure — G7 and UNESCO efforts on social/financial AI disclosure requirements

Investment Allocation Categories (Not Yet Standardized)

Frontier model development (very high funding)
Data infrastructure and training (very high funding)
Safety infrastructure and testing (underfunded)
Open-source tooling (underfunded)
Standards and certification (underfunded)
Global South AI adoption (severely underfunded)
Human capital / skills development (underfunded)

Regulatory & Policy Mechanisms Discussed

Public procurement conditionality — Requiring AI safety investment as condition of government contracts
Tax deductibility — Making safety R&D expenses tax-deductible to incentivize corporate investment
Mandatory disclosure — Requiring companies to report AI safety investments and outcomes
Personal legal liability — Making executives criminally responsible for AI harms (Spain model cited)
Sovereign wealth fund investment — Direct public capital allocation to trustworthy AI companies

Document Quality Note: The transcript contains significant audio transcription artifacts (repetitions, unclear speaker attributions, audio corruption). This summary prioritizes extracting verifiable conceptual content while flagging where speaker attribution is uncertain. The core arguments and evidence presented are accurately reflected.