U.S. Department of War and Anthropic Dispute on AI Safeguards

As governments increasingly deploy Artificial Intelligence (AI) in areas such as public administration, national security, and policymaking, questions about its safe usage and accountability become central.

• The Pentagon vs. Anthropic incident highlighted tensions between government AI deployment and company control over AI systems.
• Key question: Who governs the systems that govern us?

Potential and Caution in AI Deployment

Raman Jit Singh Chima notes the potential transformative effects of AI, such as improved data access and analysis.

• AI can strengthen state capacity depending on the problem addressed.
• Risks include misuse in facial recognition and surveillance, requiring a 'do no harm' approach.

Isha Suri emphasizes AI's effectiveness in well-scoped use cases, like distinguishing lung infections during COVID-19.

• Governments need clarity on objectives before adopting AI systems.
• Consider necessity and proportionality tests for AI deployment.

Privacy Concerns and Data Management

Isha Suri questions the notion of efficiency and highlights weak evidence on productivity gains from AI.

• Concerns about labor substitution and lack of transparency in data usage.
• Importance of informed consent and safeguarding data privacy.

Raman Jit Singh Chima argues against the assumption that more data leads to better AI, citing inefficiencies and risks.

• Alternatives include smaller models and on-device AI systems.

Government and Private Sector Partnerships

Discussion on whether public datasets should be treated as strategic assets or shared with private companies.

• Potential risks involve privacy, security, and sovereignty.
• Concern over public resources enabling private value extraction.

AI and Public Interest

Considerations for governments in adopting AI, focusing on public interest and democratic values.

• Learning from past digital infrastructure projects to avoid costly and inflexible arrangements.
• Examining trade-offs in welfare delivery and accountability gaps in public-private partnerships.

Global Adoption and India's Position

Debate on whether India should adopt AI based on global trends or define its own objectives.

• The need for building foundational scientific capacity for technological sovereignty.
• Concerns about dependency on foreign technology and potential sovereignty risks.

Conclusion and Recommendations

Raman Jit Singh Chima advises caution in AI deployment, focusing on clear use cases and long-term sustainability.

Isha Suri stresses the need for governments to clearly define objectives and assess risks vs. benefits before adopting AI.