Context: An Anthropic threat-intelligence report covering December 2025–August 2026 documented multiple categories of artificial intelligence (AI) misuse, highlighting the growing sophistication and scale of AI-enabled harm.
- Call for Slowdown: Anthropic CEO Dario Amodei called for slowing frontier AI development, with other leading AI figures subsequently expressing support. The debate highlights the need to strengthen AI governance alongside AI development.
Emerging Nature of AI Misuse
- AI as an Orchestration Layer: AI is increasingly capable of coordinating multiple software systems and executing multi-stage operations, rather than merely generating harmful content.
- “Uplift” Effect: AI can increase the speed, scale and sophistication of harm, enabling fewer individuals with limited expertise to conduct complex operations.
- Disinformation Example: An alleged Bangladesh operation reportedly used AI-generated content, automated video creation and scheduling tools to operate 29 accounts and produce around 1,500 fabricated stories, illustrating large-scale automated influence operations.
- Biological Risks: Anthropic acknowledged that its most capable models may provide meaningful assistance to sophisticated actors conducting biological weapons research, highlighting limitations in existing safeguards.
Key AI Governance Risks
1. Automated Disinformation
- AI can generate multilingual content rapidly and target specific demographic groups, making political misinformation easier to scale and harder to detect.
- This is particularly significant for India because of its large, linguistically diverse electorate and continuous electoral cycle.
2. AI-Enabled Intellectual Property Theft
- The reported Alibaba-related distillation campaign involved 151 million AI exchanges aimed at replicating a competitor's capabilities, highlighting risks from systematic model extraction and fraudulent API use.
3. Surveillance and Privacy
- AI-enabled surveillance can transform technology into an instrument of mass monitoring and social control, raising concerns regarding the right to privacy under Article 21.
Limitations of Voluntary Governance
- Competitive Pressures: Unilateral development slowdowns are difficult to sustain because of commercial incentives, large capital investments and geopolitical competition, particularly amid strategic competition with China.
- Regulatory Gap: The rapid expansion of frontier and agentic AI is creating risks faster than existing governance frameworks can adapt.
- Need for Binding Rules: Voluntary disclosure and industry self-regulation may be insufficient where AI systems can cause large-scale, automated harm.
Existing Indian Framework
- Information Technology Rules, 2021: India established legal obligations for significant digital platforms under the broader principle of safe, trusted and accountable digital platforms.
- Platform Accountability: The experience of regulating large digital platforms provides a foundation for extending accountability mechanisms to increasingly autonomous AI systems.
Proposed AI Guardrails for India
- Mandatory Misuse Reporting: AI platforms above a defined scale should report significant detected misuse to Indian Computer Emergency Response Team (CERT-In) and a designated AI safety authority.
- AI Content Provenance: Mandatory watermarking or provenance mechanisms for AI-generated political and public-interest content could help identify synthetic media.
- Agentic AI Regulation: Platform accountability rules should explicitly cover agentic AI, which can independently execute actions across software systems.
- Prevent Model Distillation Abuse: Legislation should prohibit systematic unauthorised model distillation and fraudulent mass API access.
- Statutory AI Safety Authority: A dedicated authority could be empowered to compel disclosures, conduct audits and impose proportionate restrictions on high-risk AI systems.
What is Agentic AI?
- Definition: Agentic AI refers to systems capable of planning, making decisions and taking actions through connected digital tools, rather than merely responding to user prompts.
- Governance Significance: Such systems require stronger safeguards because potential harm can arise from autonomous execution and interaction with external systems.
Way Forward
- Governance Alongside Innovation: AI development should be accompanied by equally rapid development of legal, institutional and technical safeguards.
- Risk-Based Regulation: Rules should focus particularly on high-risk applications such as autonomous agents, election-related manipulation, surveillance and biological applications.
- Auditable AI: High-capability systems should have mechanisms for traceability, incident reporting, independent audits and accountability.
- International Cooperation: Since AI risks cross national borders, India should participate in global efforts to establish interoperable safety and accountability standards.
- Rights-Centred Framework: AI governance should balance technological innovation with privacy, democratic integrity, security and individual rights.
