The big headlines out of last month’s Paris AI Summit missed an even bigger story: responsible AI is, indeed, possible, and the largest venture capital limited partners are lining up to make it a reality.
Mainstream headlines focused on how much money France and its allies were putting down to fund data centers and chips; and Vance’s totally off speech accusing Europe of anti-competitive regulation and unnecessary worries about safety.
At private meetings and public panels, big pocketed asset owners, including European behemoths European Investment Fund and BPI France, as well as US-based Omidyar Network and StepStone, were taking a different approach: Expressing concern about the possible dangers of the new technology and confident that – with the right expertise – they could fund a responsible version of it.
“Funding responsible technology will be key to unlocking growth and mitigating the risks associated with deploying cutting edge technology,” said Suzanne Tavill of StepStone.
In a private meeting with big-pocket philanthropists, Joshua Bengio, the world’s most cited artificial-intelligence researcher, made a plea: use those hundreds of millions to contribute to a different kind of safe and secure AI.
We need actually open alternatives to the not-so-open Open.AI and Mistral; we need to start thinking about safe and secure AI applications, from generative and agentic to supportive tools. Bengio felt that between his native Canada and Europe, this alternative was still possible – at least if significant action was taken within a two-year window. ROOST, the new non-profit backed by Eric Schmidt’s philanthropies, held court about its free, open-source safety tools to public and private organizations of all sizes across the globe.
“AI is the visible peak of a much deeper technology iceberg,” said Paul Fehlinger of Project Liberty Institute, who co-convened the first transatlantic LP process on Responsible Investment in Data and AI with VentureESG and is also supporting ROOST. “The fundamental question is what kind of data economy we want to build now. This will affect everything from GDP growth to innovation capacity and entrepreneurship worldwide for decades to come. LPs sit at the top of the pyramid and have an outsized impact on which technologies will scale.”
In Paris, the mood from the big money and big brains —the asset owners and philanthropists, as well as academics – thinking about responsible AI were reflective of the research VentureESG did with a much wider group of the biggest institutional asset owners and asset managers over recent months.
Responsible investing in responsible AI
Over the past few months, we’ve spoken to 26 of the largest institutional asset owners investing as LPs in venture capital globally to find out their perspectives and practices on responsible investing.
Not one of the big pension funds, state funds, including European Investment Fund and BPI France, who also joined our launch in Paris, nor the asset managers like StepStone, suggested that they are going to do less on responsible investing than they were two years ago and most are planning to do even more, despite some politicised backlash.
Why? Simple: it makes good business sense to think broadly about risk and opportunity — and that means deepening processes to address material issues, even if this means being less open publicly. And the same approach applies to investments in AI: Three out of four LPs we spoke to recognise specific significant long-term investment risk in the negative externalities of data and AI technology.
However, concerningly, only 19% of respondents to our survey felt that they had sufficient internal expertise to develop specific responsible investing approaches for data and AI, compared to 75% for ESG in general.
Developing an understanding of data and AI — both in terms of opportunities and risks — was almost universally identified as an area where improvement was urgently needed.
Tools and expertise
While eager to learn, LPs don’t quite know where to look when it comes to responsible AI; available frameworks are not specific to startups and VC. Research is often muddled and complex – starting with convoluted academic language which needs translating for practitioners. Worryingly, LPs at the moment tend to rely on the knowledge and expertise of VC managers but recognise they are unable to sufficiently evaluate their GPs.
Regulation can play an important role here – if it avoids being clunky. In setting the guardrails of safe, open, and secure AI, light-touch rules can create a framework for investors to think about responsible AI without hampering innovation.
Like what happened in Europe with the more general sustainability regulation SFDR, the right kind of AI regulation can lock in high-level guardrails, while allowing the LPs to translate them into asset-class specific guidance. In the absence of stringent regulation, LPs should do the work collaboratively – and conduct strict due diligence nevertheless.
In fact, addressing specific AI risks such as bias, data licensing and privacy, misinformation, and hallucinations will help to create more resilient AI business models.
Have you ever encountered someone who wouldn’t want powerful and safe AI? So far, we just haven’t really proposed that option. We have strong reason to believe that the pursuit of a more responsible vision of AI will lead to a broader, more satisfied customer base.
Win (LP) – win (GP) – win (AI company) – win (customer) – win (society).
Johannes Lenhard is the CEO of VentureESG and affiliated lecturer at the University of Cambridge.
Oliver Nixon is the research lead at VentureESG
