Open-weight artificial intelligence can widen access while complicating accountability. Analyse the opportunities and risks for India, and suggest a governance approach.

Open-weight models make trained parameters downloadable and adaptable. They can democratise experimentation, but openness does not automatically disclose data, code or safety practices.

Opportunities for India

• Indian firms can adapt models to local languages, run them on domestic infrastructure and reduce dependence on foreign APIs.
• Competition among models may lower costs and encourage hardware-neutral innovation.

Accountability gaps

• Weights can be modified, safeguards removed and misuse become harder to attribute.
• Autonomous agents can operate beyond intended constraints; existing legal boundaries may leave gaps in responsibility.

Governance path

1. Require risk documentation, provenance and incident reporting for high-impact deployments.
2. Use proportionate licensing, independent testing and privacy-by-design rather than blanket prohibition.
3. Invest in Indian-language evaluation, secure compute and public-interest applications.

open-weight model; AI safety; accountability; provenance; privacy; autonomous agents; Indian languages; proportionate regulation.