Systemic Credit Risks Emerge as AI-Driven Lending Exacerbates Financial Inequality and Market Instability

Indigenous financial systems, such as the Māori *whanaungatanga*-based lending models in New Zealand or the *tanda* systems in Mexico, prioritise collective well-being over individual credit scores. These systems treat debt as a social contract rather than a financial instrument, contrasting sharply with AI-driven lending’s focus on individual risk profiles. However, these models are systematically excluded from mainstream financial infrastructure, reinforcing the dominance of extractive AI systems.

The rise of AI in lending mirrors historical patterns of financial innovation being weaponised against marginalised groups, from 19th-century pawnbrokers to 20th-century redlining. Algorithmic lending echoes the discriminatory practices of the Home Owners’ Loan Corporation (HOLC), which used biased data to deny mortgages to Black communities. The current AI-driven credit scoring systems, trained on datasets riddled with historical biases, risk repeating these failures at scale, with automated decisions that are harder to challenge.

In Japan, *shinise* (trusted long-term lenders) rely on relational trust rather than credit scores, a model that has sustained small businesses for centuries. African ROSCAs and Latin American *tandas* demonstrate how communal risk-sharing can outperform individualised lending in resilience. However, these systems are often sidelined in favour of AI-driven models that favour scalability over trust, despite evidence that relational lending reduces default rates in low-income communities.

Studies show AI credit models exhibit racial and gender bias, with Black and Latino borrowers up to 30% more likely to be denied loans by algorithmic systems than white borrowers with identical profiles. Research from the Federal Reserve indicates that AI-driven lending exacerbates procyclicality, amplifying economic downturns by tightening credit during crises. The lack of transparency in these models—often treated as 'black boxes'—violates basic scientific principles of reproducibility and accountability.

Artistic and spiritual traditions often frame debt as a moral and spiritual burden, as seen in Dante’s *Inferno* or Buddhist teachings on attachment to wealth. Modern AI lending systems strip away this moral dimension, reducing human relationships to numerical scores. Indigenous and Afro-diasporic traditions, such as the Yoruba concept of *àṣẹ* (sacred power), view financial transactions as part of a larger cosmic balance, a perspective absent in algorithmic lending frameworks.

Scenario modelling suggests that unchecked AI lending could trigger cascading defaults in a future economic downturn, particularly if models are trained on data from a single economic cycle. Alternative futures include decentralised finance (DeFi) models that use blockchain for transparent, community-controlled lending, or public credit registries that prioritise fairness over profit. The long-term stability of financial systems may depend on shifting from extractive AI to regenerative models that account for social and ecological costs.

Marginalised borrowers—particularly Black women, Indigenous communities, and low-income gig workers—are disproportionately harmed by AI lending, as their data is often excluded from training datasets or misclassified as high-risk. The voices of these borrowers are systematically excluded from policy discussions, with lenders and regulators framing their exclusion as a 'risk management' issue rather than a structural injustice. Grassroots organisations like the Debt Collective and the Center for Responsible Lending have documented how algorithmic lending deepens racial and economic divides, yet their insights are rarely centred in mainstream analyses.