Impact investing and VC screening criteria on AI

Movement organisations advocate for and help draft AI-specific screening criteria that impact investors, socially-responsible-investment funds, and venture capital funds apply to incoming investment decisions — the moment when capital first enters an AI startup or the AI arm of an existing firm. The vehicle is a published criteria framework (ICCR-style investor guidance, PRI-style principles, IRIS+ metric extensions) that individual funds, pension funds, and endowments adopt as portfolio-construction rules: no equity into companies whose named practices include developing autonomous weapons, biometric mass surveillance, unlicensed generative-AI training on copyrighted work, or work with contracted state actors on named enforcement operations. The strategy converts the socially-responsible-investment infrastructure the movement has already built for climate and human-rights concerns into an AI-specific screening layer.

An actor chooses this strategy because it works on the primary market — the funding round, the term sheet, the fund's screening committee — rather than the secondary market of already-public equity where shareholder resolutions and divestment operate. A criteria framework adopted by a fund excludes a category of AI investment before the capital is committed, at a point in the company's lifecycle when access to the impact-investor pool is a distinct and sometimes decisive source of capital. The strategy also produces a movement-legible artifact — the criteria document — that becomes a rallying point for the next fund and the reference the next campaign cites, compounding across the sector rather than fund-by-fund. Established SRI and impact-investing infrastructure (ICCR, PRI, US SIF, Global Impact Investing Network) provides the delivery channel; the movement's job is to author the AI-specific criteria and organise adoption.

It trades reach into the largest capital pools for coherence of criteria. The frontier-AI labs and the largest deployers raise capital predominantly from investors who are not impact-screened — sovereign wealth funds, mainstream VC, corporate strategic capital, private equity — and the impact-investor pool is a small fraction of AI capital by volume. A criteria framework adopted comprehensively across the impact-investor sector might still leave the target vendor's next funding round oversubscribed by unscreened capital. The strategy is also vulnerable to definitional gaming: criteria that name specific practices are evaded by companies that reorganise their disclosure to avoid the criterion, and criteria that are principle-based rather than practice-specific dilute to the point of admitting the target vendor under a permissive read. Sustained coalition maintenance of the criteria against these two failure modes is what separates a real screening framework from a nominal one.

Ecology

Distinct from shareholder resolutions on AI vendors. Shareholder resolutions engage existing holdings — the investor is already inside the company and using its voting rights to advocate change. This strategy operates upstream of that position — the investor uses its screening criteria to refuse entry into the company in the first place. A movement that runs both arms operates on two temporal channels of the same investor's relationship with the target company: refusal at entry, engagement at exit.

Distinct from endowment and pension divestment from AI. Divestment exits existing positions in named vendors; screening criteria refuse entry to a category of investment defined by practice. Divestment is target-specific and campaign-driven; screening criteria are practice-specific and framework-driven. The two combine into a full-cycle refusal — screening prevents entry, divestment exits legacy positions — but the organising forms differ substantially: divestment is a campus/pension-member mobilisation, criteria adoption is a fund-governance intervention.

Distinct from institutional procurement refusal of AI vendors. Procurement refusal targets the institution's decision to buy an AI product; screening criteria target the institution's decision to invest capital into an AI vendor. Both are movement-pressure on the institution's operational choices; they differ in whether the institution's role is customer or investor, and in which department (procurement vs. treasury) becomes the movement's inside-institution interlocutor.