What are the ethical concerns in biological and socio-cultural anthropology because of recent advances in AI and genetic research?

Biological anthropology

Human genetics, ancient DNA, and medical genomics need informed consent that a village can actually give. A signed form in English is not consent among a PVTG. Community consent (after Nuu-chah-nulth, Havasupai, and Indian debates on tribal sampling) sits beside individual consent because a genome is never only one person’s secret: relatives and future children are implicated. Benefit-sharing and the spirit of the Nagoya Protocol require that if a community’s genetic or associated traditional knowledge is used, the community is a partner, not a raw-material site.

• Privacy and re-identification: even “anonymised” genomic files can be linked. Databases must limit access. Ancestry tests sold to the public can freeze colonial racial types (Risley, Guha’s older taxonomy) as commercial identity. Anthropologists must not treat a PCA plot as a caste origin certificate. Rakhigarhi aDNA is a scientific result; using it to police who is “indigenous” is a political misuse.

• Medical genetics: counselling, not coercion. Sickle-cell and other screening in tribal belts is ethical when it is voluntary, confidential, and tied to care. It is unethical when it becomes a condition of marriage or of a government benefit. Human genome editing for heritable change is under international moratoria and Indian review (ICMR, Department of Biotechnology, ethics boards) because of uncertain harm to descendants and because enhancement could deepen class inequality. This is a governance issue. It is not a recipe.

• Collections: skeletal and blood archives from colonial and ASI contexts need repatriation or restricted use when communities object. The Helsinki tradition, CIOMS, and ICMR National Ethical Guidelines are the Indian working texts.

Socio-cultural anthropology

Artificial intelligence here means machine systems that classify images, speech, text and behaviour. Ethnographic fieldnotes, photographs and interview recordings, if scraped or donated without permission, can train models that later identify activists, migrants or minorities. Anonymity, a classic field ethic (AAA, ASA, Indian university IRBs), is harder when faces and voices are searchable.

Biased training data reproduce caste, gender and racial stereotype as if they were nature. Predictive policing and welfare-fraud scoring can punish the poor for being poor. Anthropologists who advise such systems share responsibility. Deepfakes of ritual or of a person damage reputation and sacred secrecy (some Australian and Indian communities already restrict images of the dead).

• Authorship and labour: students and informants who clean data for a model must be credited and paid. A chatbot that answers as if it were a tribe’s elder is a representation harm.

Shared duties

Do not collect DNA or digital traces because it is possible. State a research question, a storage limit, a destruction date, and a community veto. Refuse projects whose purpose is covert surveillance or the ranking of peoples. The Ghurye–Elwin lesson still applies: knowledge that only serves the state against a vulnerable group is not autonomous anthropology.

National Digital Health and Aadhaar-linked records raise the same consent problem at population scale. Anthropology’s contribution is to insist that groups, not only individuals, can be harmed.