Investing with AI, investing in AI: A dual lens for due diligence and impact

AI is positioned to account for more than half of all VC money invested in 2025. Yet while capital is flowing at scale, frameworks to guide responsible and impactful investment are still nascent. 

A survey launched at the Paris AI Action Summit in February 2025 as part of the first LP process on responsible investment in AI and data technologies, convened by ReframeVenture, Project Liberty Institute and Omidyar Network, that now engages more than 60 institutional LPs with over $5 trillion in capital, found that three out of four LPs are concerned about the systemic risks AI poses across their GP portfolios. 

Adapted due diligence is no longer optional. It is urgent.

To respond, ReframeVenture, ImpactVC, and Project Liberty Institute launched a new complementary initiative at SuperVenture 2025 to help GPs update frameworks and develop practical tools for responsible and impactful AI investment. Momentum is building: across conversations with LPs in North America and Europe, expectations are rising. GPs will need credible approaches to diligence that protect returns, meet fiduciary duties, and build trust in the markets they help shape.

At recent investor gatherings I participated in together with partners, one challenge stood out. Investors often conflate two distinct ideas: investing in responsible AI itself and using AI to drive impact in other sectors. Untangling the two is not semantics; it is the difference between managing risk and shaping markets. In that context, the dual lens, “investing with AI” and “investing in AI” is a way to clarify the conversation.

Impact investing with AI

Investing with AI, where AI amplifies existing theses across sectors like health, climate or education to do financially well and good for societies, the planet, and markets, demands new AI-specific due diligence and risk frameworks. These protect returns and strengthen theses, ensuring fiduciary duties are met while portfolios remain resilient. The thesis is not new; the diligence is.

Just as cloud computing underpinned the last wave of startups, AI is now embedded across verticals, from digital health and climate tech to financial services. A startup deploying drones to cut pesticide use may apply AI to identify pests more precisely; another may accelerate drug discovery by identifying promising compounds faster. In both cases, the impact thesis (e.g. sustainable agriculture or eradicating cancer) remains unchanged, but AI acts as a powerful new catalyst.

Yet new infrastructure brings new risks. Once AI becomes part of a company’s products or operations, the scope of due diligence expands to issues such as model provenance, data governance, autonomy, oversight, safety, explainability and regulatory compliance. Central to all of this is data agency, ensuring individuals and enterprises have real control over the data behind AI systems, from the training sets used to build them to the contextual data shaping outputs every day.

AI due diligence remains complex. Investors are only beginning to grasp the distinction between AI’s effects on financial performance and its broader societal and market impacts. Unlike the relatively linear goal of achieving net zero, AI’s multi-dimensional nature demands new indicators, frameworks, and governance approaches. Despite the abundance of resources from the 298-page AI Action Summit Safety Report to the more high-level OECD AI Principles, or the Fair Data Economy Task Force with members such as Economics Nobel Prize winner Daron Acemoglu, investors still struggle to translate them into practical tools for investment committees.

Some still question whether responsible AI is financially material. But the direction of travel across policy and markets suggests it is becoming hard to ignore. These issues go straight to valuation and exit potential: can you underwrite a Series A in digital health if the provenance of the training data is unclear? Do you back an edtech platform whose “agentic” tutors act unpredictably? Responsible AI is not a side issue; it is material risk management requiring adaptive frameworks that evolve as large-scale adoption surfaces new unintended effects.

Impact investing in AI

If investing with AI demands sharper diligence, investing in responsible AI opens an entirely new vertical that defines the system itself. As AI powers the infrastructure on which all other digital innovation depends, the underlying choices about safety, accountability, fairness, and data agency ripple through every sector.

This is not theoretical. Leading researchers, including AI “godfather” Yoshua Bengio, warn that humanity has only a narrow window to align AI with human agency. The rise of agentic AI, i.e. systems capable of planning and executing actions independently, raises the stakes further.

For investors, this is where impact and returns increasingly converge. Investing in AI today is not just about chasing the next growth cycle. It’s about building the foundations of a more trustworthy and resilient digital economy. Opportunities cut across the stack: infrastructure and protocols that secure data provenance and user control; assurance and safety tools that make systems transparent and reliable; and applications that prove fair, sustainable data models can scale. Taken together, these areas form a clear emerging thesis linking long-term value creation with accountability and human agency.

Capital in this space does more than mitigate risk. As investors who saw the regtech boom after the financial crisis or the rise of the cybersecurity market will recall, it accelerates adoption, shapes standards, and builds the rails for markets designed to endure. Companies embedding transparency and accountability gain competitive advantage such as winning customers, navigating regulation smoothly, and sustaining stronger valuations. Early investors in responsible AI infrastructure will not only capture upside but also help set the terms of the next growth cycle.

Why the dual lens matters

Both lenses are essential. Investing with AI strengthens existing theses as the technology permeates every sector. Investing in AI addresses systemic risks and builds the infrastructure for a more resilient digital economy.

The lesson for investors is threefold. First, AI is not a neutral enabler. It changes the risk calculus in every sector. Second, treating AI as a vertical opens a new impact thesis, positioning infrastructure, safety and agency as investable markets. And third, the distinction between “with AI” and “in AI” is not semantics. It is the difference between managing near-term risks and shaping long-term markets.

For LPs and GPs committed to both performance and impact, this dual lens is pragmatic. Responsible AI is not a cost center. It is the foundation for market trust, portfolio resilience, and durable value creation. Those who master both lenses won’t just invest in the AI economy. They’ll define its trajectory.

Paul Fehlinger is the director of policy, governance innovation and impact at Project Liberty Institute, a VC-focused affiliate at Harvard, and the publisher of VCandPolicy.com. 

Guest posts on ImpactAlpha represent the opinions of their authors and do not necessarily reflect the views of ImpactAlpha.