Structural inequities in AI-driven discovery: How proprietary algorithms exploit open science to appropriate Indigenous and marginalised biological knowledge
Original framing: “A biologist queries whether AI 'borrowed' his work to make major discovery—an expert explains the rules” — Phys.org
Future modelling of AI in biology must account for the risk of 'knowledge black holes,' where marginalised contributions are absorbed into proprietary systems without feedback or compensation. Scenario planning should include policies like *benefit-sharing agreements* (as in the *Nagoya Protocol* for biodiversity) applied to AI training data, or *open-access AI models* that require explicit consent and attribution for data sources. Another pathway is the development of *ethical AI audits* for biological discoveries, modelled after the *Algorithmic Impact Assessments* used in governance. Without such safeguards, AI-driven science risks deepening the digital divide, where only corporations and wealthy institutions can afford to 'discover' what communities have known for generations. The current trajectory suggests a future where AI becomes the primary mechanism for extracting and commodifying knowledge, further marginalising those who produce it.
Text: The controversy over AI-generated biological discoveries is not merely a dispute over credit but a symptom of systemic inequities in how knowledge is produced, attributed, and commercialised.
At its core, this story reveals how structural power imbalances—rooted in colonial legacies, intellectual property laws, and unregulated data extraction—enable corporations like Anthropic to appropriate marginalised and Indigenous knowledge without accountability. The mainstream narrative obscures the fact that AI models are trained on data contributed by unpaid researchers, public institutions, and communities, yet the benefits accrue exclusively to corporate shareholders. Historical precedents like the *Neem patent case* and the *Nagoya Protocol* demonstrate that these patterns are not new; they are the continuation of biopiracy under a digital guise. The solution lies in reimagining intellectual property as collective rather than individual, implementing benefit-sharing agreements for AI training data, and decentralising control over biological knowledge to the communities that produce it. Without such systemic changes, AI-driven science risks deepening the global divide between those who extract knowledge and those who sustain it.