Amnesty International says India AI Impact Summit failed to rein in harmful practices of governments and tech firms
Amnesty International said the India AI Impact Summit did not deliver binding rights protections or meaningful participation for affected communities.
Summary
Amnesty International has issued a statement responding to the close of the AI Impact Summit in New Delhi, India, which ran from 16 to 20 February. The reaction came from Erika Guevara Rosas, one of the organisation's senior directors, who judged that the summit did not curb harmful uses of artificial intelligence and that its tone clashed with conditions on the ground in India.
According to Amnesty, AI systems in India are driving a state-led push toward authoritarian practices that strengthens state and corporate control. Guevara Rosas said civic space is narrowing at an unprecedented pace and that marginalized communities affected by these systems face ongoing harm. She argued that world leaders praised India's technological advances while human rights concerns were played down. Amnesty points to its own research showing that facial recognition and public-sector automation have endangered privacy and social protection and led to exclusion and discrimination, while mass surveillance systems are expanding.
The statement criticises the summit's emphasis on sovereignty, innovation and democratisation, which it says turns AI into a contest of accumulating power and economic growth. Amnesty says civil society and affected communities were barely involved. More broadly, it contends that AI summits have not advanced needed regulation, have been largely ineffective at delivering binding rights protections, and have favoured technology-centred narratives and soft governance tools, while the distance grows between state action and an increasingly unchecked industry.
Amnesty calls on states to change course, adopt binding guardrails that prohibit technologies incompatible with human rights, and create genuine channels for public participation. As background it cites its 2024 documentation of an automated system in Telangana excluding thousands from social protection, and its 2021 study of facial recognition in Hyderabad.