Trustworthy AI Investments : Capital Allocations as AI Governance

Abstract

The panel examined how the flow of capital—venture‑funding, sovereign wealth, philanthropic grants and public procurement—acts as a decisive lever for AI governance. Panelists highlighted the stark imbalance between massive investments in frontier model development and the meagre resources allocated to safety, openness and inclusion. They explored why this gap exists (competitive pressure, national‑security dynamics, short‑term return expectations), how standards, certification programmes and public‑policy incentives can rebalance the stack, and what role governments, investors, standards bodies and end‑users must play—especially in regions that lack sufficient AI‑related funding. The discussion concluded with concrete ideas for how a billion‑ or multi‑billion‑dollar fund could be deployed to accelerate trustworthy AI across the whole ecosystem.

Detailed Summary

• Sophie introduced the panel, noting that conversations about AI governance and AI investment have largely proceeded in parallel silos.
• She argued that governance is often seen as “boring” while investment is “cool”, a mismatch that the session aims to correct.
• The goal: view capital allocation as a tool—a lever that shapes which AI systems are built, who benefits, and how safety, inclusion and resilience are prioritized.

2. Capital as a Governance Lever (Amir Banifatemi)

• Amir outlined the classic AI‑governance toolbox:
  • Soft tools – ethical frameworks, voluntary guidelines.
  • Hard tools – laws, regulations.
• He emphasized that capital flows ultimately decide which models are created and whose interests they serve.
• The panel would use new investment data (from the Global Partnership on AI and Mozilla Ventures) to illustrate critical funding gaps in safety infrastructure, foundational research, and “global‑majority” innovation.

3. Investment Landscape: Where the Money Goes

3.1 Massive Funding for Frontier Models

• Mohamed Nanabhay described the “planet‑scale” scale of AI investment: billions of dollars flowing into model training, data acquisition, and data‑center infrastructure.
• He likened the current wave to the early Internet era—rapid growth with governance lagging behind.

3.2 Negligible Funding for Safety & Open‑Source Tools

• Mohamed noted that only a tiny fraction of capital reaches safety‑related tooling, open‑source ecosystems, or governance‑centric projects.
• Competitive pressure among a handful of “frontier labs” (e.g., OpenAI, Anthropic) drives the focus on speed to market rather than risk mitigation.

3.3 Competitive & Geopolitical Drivers

• He highlighted national‑security concerns (U.S.–China AI rivalry) that push governments and private actors to prioritize capability over safety.

4. Standards, Certification & Public‑Sector Levers (Alpesh (Al Pasha))

• Alpesh explained that capital allocation intersects with both liquidity (investment dollars) and human capital (skilled talent).
• Standardisation serves three functions:
  1. Provides a public‑utility baseline that lower‑resource actors can adopt.
  2. Incentivises cross‑border alliances that pool capital for common standards.
  3. Enables ecosystem formation where certain groups own data, frameworks or test‑beds.
• He cited his organisation’s AI‑ethics certification programme, now used by several governments and firms to demonstrate responsible AI practices and to unlock capacity‑building funding.

5. Government Perspective: Singapore’s Approach (Wan Sie Lee)

• Lee described Singapore’s policy mix to foster trustworthy AI without directly funding the big labs:
  • Startup grants, tax exemptions, and public‑policy‑driven venture funding to nurture AI applications in education, finance, health, etc.
  • Creation of “Lorong AI”, a physical co‑working hub that encourages community learning and rapid prototyping.
• In AI governance, Lee outlined two complementary levers:
  1. Third‑party assurance firms (e.g., Rosaro, funded by sovereign fund Temasek) that provide testing, certification and risk‑assessment services.
  2. Certification signals that can be required by venture capitalists or procurement officers before they invest.

6. Funding Asymmetries Between Global North & South

6.1 North‑Centric Investment Patterns

• Julien Billot (Scale AI) argued that market incentives already reward trust: companies cannot secure funding or board approval unless AI solutions are demonstrably trustworthy.
• He likened AI to financial‑market auditing—future AI products will need certified compliance, much like accounting standards.

6.2 South‑Centric Gaps & Climate‑Finance Analogy

• Gabriela Ramos compared the current AI‑funding gap to early climate‑finance: only when regulatory incentives, public procurement rules, and dedicated grant programmes were introduced did capital start flowing to climate mitigation.
• She advocated for similar policy tools (tax credits, procurement mandates, dedicated $100 bn‑type funds) to direct capital toward AI safety and inclusive innovation in developing economies.

7. Why Is Investment in Safety Lagging?

• Panelists identified several structural causes:

  1. Speed of capital vs. speed of regulation – investors move faster than lawmakers can enact AI‑specific statutes (e.g., EU AI Act is still pending).
  2. Short‑term EPS pressure – publicly listed firms focus on quarterly earnings, making safety‑related spend appear as a non‑core cost.
  3. Market ignorance – many limited partners and sovereign funds lack awareness of AI‑risk metrics, so they do not pressure downstream VCs.
  4. Consumer indifference – unlike aviation, most users do not demand safety certifications for AI services, limiting downstream market pressure.

• Bottom‑up vs. Top‑down: Both cultural change within organisations (building a “safety‑by‑design” mindset) and regulatory mandates (e.g., required certifications) are needed.

8. Role of Consumers, Boards & Individual Actors

• Mohamed (Mozilla) reminded that end‑users can drive demand for trustworthy products, as Mozilla’s 25‑year history shows a viable business model balancing privacy, profit and public interest.
• Gabriela underscored that board members could be held legally responsible for AI‑related harms (citing Spain’s upcoming liability rules), creating powerful internal incentives for governance.
• Sophie asked the audience to reflect on personal incentives: up‑skilling, demanding transparency, and participating in public‑consultations are ways individuals can influence capital flows.

9. Policy Instruments to Shift Capital

Instrument	Example / Panel Insight
Public procurement with safety clauses	Singapore’s IMDA accreditation stamp could become a procurement prerequisite, reducing friction for trustworthy AI vendors.
Impact‑venture capital	Funds that require dual bottom‑line (financial return + societal benefit) to back AI‑trust startups (e.g., Mozilla Ventures, Scale AI).
Tax incentives / deductions	Governments could allow companies to deduct AI‑safety R&D expenses, echoing climate‑finance mechanisms.
Certification & standard‑setting bodies	IEEE, OECD, and national agencies can create interoperable benchmarks that become de‑facto market requirements.
Sovereign catalytic capital	Large sovereign wealth funds can earmark “trustworthy AI” pools to seed early‑stage safety research.

10. “If We Had a Billion (or Five) Dollars…” – Visionary Allocation

Speaker	Proposed Allocation
Alpesh (standards)	Invest across the entire AI stack (edge, cloud, on‑device) to embed trust by design; fund reusable tooling and open standards.
Wan Sie Lee (human capital)	Direct funds to education & up‑skilling of developers, policymakers and citizens, especially women and under‑represented groups.
Julien Billot (certification)	Build a global AI‑certification regime that companies can display (e.g., “AI‑Ready” label), partnering with standards bodies.
Gabriela Ramos (policy incentives)	Create a large‑scale fund for public‑procurement‑linked safety projects and support sovereign catalytic capital mechanisms.
Mohamed (open‑source & commons)	Seed public‑utility AI platforms and open‑source safety tools to lower entry barriers for the global majority.

Collectively, the panel argued that no single billion‑dollar bucket is sufficient; a portfolio approach spanning standards, human capital, certification, public procurement and open‑source ecosystems is required.

11. Closing Remarks

• The moderator (Amir) thanked participants and highlighted that money truly talks: aligning capital with trustworthy‑AI principles can accelerate the societal benefits of AI while reducing systemic risk.
• The audience was urged to leave the session with hope and actionable ideas, rather than fear, acknowledging that the AI Impact Summit’s purpose is to “turn optimism into investment”.

Key Takeaways

• Capital flows decide AI outcomes. The majority of AI funding supports frontier model development; safety, openness and inclusion receive a marginal share.
• Governance must be funded. Soft guidelines are insufficient; hard‑budget allocations (venture, sovereign, public) are required to embed trust‑by‑design across the AI stack.
• Standards & certification are market levers. Harmonised benchmarks (e.g., IEEE, OECD) can become procurement prerequisites and VC “signal” requirements, steering capital toward trustworthy solutions.
• Regional asymmetries exist. The Global North dominates AI‑investment; targeted public‑policy tools (tax credits, procurement mandates, sovereign catalytic funds) are needed to uplift the Global South.
• Competitive pressure and national‑security dynamics currently crowd out safety spending. Faster‑moving capital markets outpace the slower legislative process, leaving governance lagging.
• Bottom‑up cultural change is as vital as top‑down regulation. Companies need internal safety‑by‑design mindsets; consumers must demand transparent, trustworthy AI.
• Public procurement can de‑risk and accelerate adoption of trustworthy AI by using accreditation stamps (e.g., Singapore’s IMDA model) that certify compliance before purchase.
• Impact‑venture capital and dual‑bottom‑line funds can align profit motives with societal benefits, providing a scalable financing mechanism for AI‑trust startups.
• Human capital investment is essential. Upskilling developers, regulators, and end‑users—especially women and under‑represented groups—creates the talent pipeline needed for trustworthy AI.
• A diversified funding portfolio (standards, certification, open‑source commons, education, and policy incentives) is the most effective way to deploy a billion‑ or multi‑billion‑dollar pool to accelerate trustworthy AI worldwide.

---

See Also:

• shaping-the-ai-narrative-trust-outcomes-and-responsibility
• boosting-ai-adoption-in-the-global-south-critical-role-of-trust-across-the-ai-tech-stack
• trust-and-safety-as-engines-for-growth-and-innovation
• applying-ai-how-ai-innovators-and-governments-reshape-the-public-sector
• mahaai-building-safe-secure-and-smart-governance
• responsible-ai-in-action-how-global-enterprises-are-building-trust-at-scale
• thriving-with-ai-human-potential-skills-and-opportunity
• who-watches-the-watchers-exploring-independent-verification-in-ai-governance
• peace-power-and-perspectives-a-value-based-approach-to-trustworthy-ai
• the-role-of-science-in-international-ai-governance