Deploying AI for good is nice, but not enough. Impact investors and others are mobilizing capital to build “good AI.”
This month’s sudden acceleration in public awareness of the impending AI disruptions to work, careers, stock markets and much more has added urgency to these investors’ efforts to shape the entire technology “stack” underlying artificial intelligence. Call it “impact in the loop.”
“The question for investors right now is whether we will fund an AI economy designed merely to maximize profits for a few, or one built to reinforce human ingenuity, agency, creativity, and the public good,” said Katy Knight of the Siegel Family Endowment, a New York-based foundation with more than $500 million in assets. Siegel backs organizations shaping the impact of technology on society and is supporting ImpactAlpha’s coverage of good AI, “Shaping the Algorithm.”
“‘Good AI’ isn’t just a set of principles,” Knight told ImpactAlpha. “It requires capital, infrastructure, and governance that ensure people and communities have real power in how these systems shape their lives.”
Knight will join this week’s Agents of Impact Call to help sketch emerging investment strategies for building AI infrastructure optimized for public benefit and positive social outcomes.
The India AI Impact Summit, kicking off in New Delhi, is a showcase for the market demand for more responsible, ethical, and increasingly, impactful, AI.
“This is not about investing with AI—using AI tools to optimize outcomes in sectors such as climate, health, or education,” says Paul Fehlinger of Project Liberty Institute, a nonprofit focused on building a people-centered digital future, who will also join The Call.
“It is about investing in the AI stack itself, treating AI as digital infrastructure whose design shapes value creation, information flows, and decision-making.”
Most obvious is AI safety, which includes assurance, guardrails and governance. Beneath that is a complex layer of “orchestration,” the central nervous system that coordinates AI models, data and systems into a cohesive workflow. As large language models grow up from chatbots to autonomous agents, the standards and protocols in this orchestration layer will enable collaboration between systems, performance optimization and governance.
“We have a narrow window to shape how AI infrastructure gets built,” said Omidyar Network’s Govind Shivkumar, who leads investments into responsible tech.
“Impact investors need to stop playing at the application layer and start engaging across the full stack — from energy to compute to the token,” he told ImpactAlpha. “If you want better outcomes for society, you have to help build the infrastructure, not just ride on top of it.”
Project Alexandria
Omidyar Network, the philanthropic investment firm of eBay founders Pierre and Pam Omidyar, itself is seeking to demonstrate the investment thesis with a spinoff investment fund, codenamed Project Alexandria.
Omidyar’s Mike Kubzansky in 2024 coordinated the $7.5 million investment, with Ford Foundation and Nathan Cummings Foundation, in Anthropic, now the rising challenger to ChatGPT for AI market share (see, “With stakes in Anthropic, impact investors seek a seat at the AI table”). Kubzansky has been mobilizing other limited partners for Project Alexandria, which aims to anchor companies in AI infrastructure, assurance and alignment to create guardrails and much-needed competition and energy efficiency in the market – assuredly a growth category for enterprise customers.
“AI is arguably the most powerful force transforming capital markets in the next decade,” Omidyar Network’s Chris Jurgens tells ImpactAlpha. “If you look at the quantum of investment going into data centers, the impact on large corporate performance, and what’s going into venture capital — and if we don’t have a holistic approach to thinking about risk and impact around that, what are we doing?”
Anthropic’s release of Claude Code and Cowork earlier this month jolted the stock markets as the implications of the AI disruption come into focus. Anthropic’s relatively more cautious style and focus on enterprise-readiness has propelled its rise, and $380 billion valuation. Last week the firm secured $30 billion, ahead of an anticipated IPO later this year. In an essay last month, Anthropic’s Dario Amodei warned of “real danger” posed by risks of AI absent a mix of actions taken by companies alongside governments.
Omidyar Network has made additional investments in Groc, UC Berkeley’s Skydeck Fund and Collab Capital, among other AI plays. Alexandria will also have a policy arm, to represent the voice of investors in discussions about AI governance and regulation.
“Part of Project Alexandria is mobilizing investors to be a policy voice in Washington calling for innovation and guardrails — that we need appropriate guardrails that have a competitive AI sector,” Jurgens said. “We come from markets. We are pro-AI and pro-innovation, and we think there needs to be sensible regulatory guardrails coming from public policy.”
Jurgens likens the emerging governance structures around AI in business and finance to the corporate adoption of climate-related policies and disclosures over the past decade to account for both opportunities and risks. The nonprofit Partnership on AI is developing a standard for AI risk management and disclosure modeled on the Task Force on Climate-Related Financial Disclosures established a decade ago.
“AI needs to be a core corporate governance and responsible investing issue, because it’s a huge source of enterprise value creation and a huge source of enterprise risk. And we’re not set up to be there yet.”
Omidyar is working to mobilize pension, insurance and sovereign wealth funds to bring the weight of institutional investors to bear on the AI rollout and elevate consideration of systemic risks. The World Benchmarking Alliance’s Collective Impact Coalition for Ethical Artificial Intelligence, with more than 60 investors managing more than $11 trillion, is seeking to ensure that tech companies integrate human rights and ethical considerations.
At the governance level, Jurgens says, “We need safe and secure and unbiased models at the heart of the engine room driving everything.” An AI safety and assurance industry should provide “the seat belts and catalytic converters and stoplights that govern the core engine. That’s an impact industry.”
And finally, investment needs to move toward AI use-cases that benefit society, not just ones that optimize business processes. “The path of least resistance is to start with cost cutting. We don’t want to stop that, but we want to make sure the positive impact cases are not underutilized,” he says. “If we’re not using AI to solve disease and address climate change, then we’re missing a huge opportunity for value creation and societal benefit.”
Siegel Family Endowment and Omidyar Network, along with the Ford, MacArthur and Packard foundations, are part of the coalition of 10 philanthropies backing Humanity AI, which represents at least $500 million pledged to AI initiatives “shaped by and for the people.” Siegel has also backed Opportunity AI, a separate community of philanthropies harnessing AI for economic mobility, as well as Responsible Innovation Labs and The Center on Rural Innovation.
LP/GP
In the stampede of venture capitalists and other investors pouring billions into AI are a handful of fund managers explicitly seeking solutions to tough challenges of trustworthy AI governance, privacy preservation, misinformation and the risks posed by ever-more autonomous AI agents.
Project Liberty, with Reframe Venture and Impact VC, is surveying VCs to map who is actually investing “in AI,” not just “with AI.”
Mozilla Ventures, the investing arm of the nonprofit Mozilla Foundation launched three years ago with an initial $35 million, and has backed “trustworthy AI” companies including Adaption Labs, Credo AI and Fiddler AI.
Mozilla Ventures brought together 30 of its portfolio companies in Barcelona in November alongside the Mozilla Festival (sponsored by both Siegel Family Endowment and Omidyar Network). A collection of essays from the convening “captures the key issues facing founders who care (and are building!) about responsible tech and trustworthy AI,” Mozilla Ventures’ Mohamed Nanabhay tells ImpactAlpha.
Among the themes: “expanding who builds, showing how diverse perspectives and resource-constrained environments generate more resilient, globally-relevant technology,” says Nanabhay, who will join this week’s Agents of Impact Call.
San Francisco-based Juniper Ventures has made more than a dozen early stage investments in “founders making AI secure and beneficial for humanity.” The fund touts “assurance infrastructure” that can safeguard AI systems, audit data models for trustworthiness, and verify authenticity of digital identities and content.
The investment manager Ex/ante spun out of Schmidt Futures in 2023 and raised an inaugural fund of $33 million from backers including Eric Schmidt, Union Square Ventures and Ford Foundation. Managing partner Zoe Weinberg backs “agentic technology,” or what she calls tools that support human agency by increasing privacy, security and information integrity. Investments include Anon, Hounddog.ai and Reality Defender.
With support from Ford (an investor in ImpactAlpha), Ex/ante built out the Builder Blueprints, a library of ethical policies (privacy agreements, terms of service, etc) that entrepreneurs can use off the shelf.
The activity suggests that investor concerns over so-called “X-risk,” for the most dire predictions of AI superintelligence, are giving way to more prosaic considerations of governance and regulatory compliance. Last week, The Wall Street Journal reported that Blackstone’s Stephen Schwarzman plans to put a “substantial majority” of his nearly $50 billion fortune into his philanthropic foundation upon his death, with a focus on ensuring AI develops responsibly.
With Reframe Venture and Omiydar Network, Project Liberty has engaged more than 60 institutional investors across Europe and North America, and increasingly Southeast Asia, representing more than $6 trillion in capital to understand how LPs are framing responsible AI and what they want to see in the market.
LPs view AI as a governance/fiduciary risk issue, and are pressuring managers accordingly, says Project Liberty’s Fehlinger, but the “impact” thesis is still early. Institutions are asking for clearer, evidence-based signals of what “good/pro-human” AI looks like, and where to find credible managers and deals.
“LPs are increasingly concerned not only about individual portfolio companies, but about compounding effects across labor markets, economic agency, and the future size and diversity of investable markets themselves,” Fehlinger wrote last week. “Demand is rising for AI systems that support interoperability, pluralism, and human oversight. Not as an ethical preference, but as a condition for institutional trust, market access and economic resilience.”
