Context:

• Anthropic’s September 2026 threat-intelligence report, covering nine months of AI misuse, highlights a fundamental shift in the nature of AI risks.
• AI is no longer merely a tool for generating harmful content; it is increasingly becoming an “orchestration layer” capable of coordinating multiple stages of sophisticated operations across different software systems.
• The report’s findings, followed by Anthropic CEO Dario Amodei’s call for slowing frontier-AI development, and support from leading technology figures, have intensified the debate over AI safety and governance.
• However, the central question should not be how to slow AI development, but how to accelerate AI governance. 

From AI Assistance to AI “Uplift”:

• AI “uplift”:
  • Anthropic documented AI misuse across seven broad categories, involving both state-sponsored actors and individuals.
  • A key concept is “uplift”—the enhancement of harmful capabilities through AI in terms of speed, scale and depth.
• Case of Bangladesh:
  • The disinformation operation combined AI-generated content with automated video production, account creation and algorithmic scheduling.
  • A single individual reportedly operated 29 accounts and generated around 1,500 fabricated stories.
  • Thus, AI reduces the human expertise, manpower and time required to conduct sophisticated harmful operations.
  • Moreover, documented cases represent only the misuse detected by platforms, meaning the actual scale of attempted misuse may be considerably larger.
• Limitations of existing safeguards: Anthropic has also acknowledged that its most capable models can no longer guarantee that sophisticated actors will not receive meaningful assistance in biological-weapons research.

The Limits of Voluntary Slowdown:

• Amodei warned that the pace of frontier-AI development has accelerated sharply and that autonomous AI-agent “swarms” could potentially produce unprecedented disruption.
• However, a voluntary industry-wide slowdown faces major obstacles -
  • Intense commercial competition among AI companies.
  • Huge levels of private capital investment.
  • Geopolitical rivalry, particularly involving China.
  • The difficulty of ensuring that all major developers simultaneously accept restrictions.
• There is a potential conflict of interest - a market leader advocating a slowdown could benefit by reducing competitive pressure while protecting its existing advantage.
• Therefore, voluntary restraint cannot constitute a durable governance mechanism.

Why Platform Accountability Must Expand?

• India’s experience: The Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021, established legal responsibilities for large digital platforms.
• Lessons from Bangladesh:
  • The Bangladesh case has particular relevance for India because of its large electorate; linguistic and social diversity; extensive digital penetration; and continuous electoral cycle across States.
  • AI can make multilingual disinformation inexpensive and scalable, while detection remains difficult.
• Therefore, platforms should have enforceable responsibilities to detect, disrupt and report coordinated AI-enabled influence operations.

Emerging Risks Beyond Disinformation:

• The Alibaba distillation campaign, involving an enormous number of AI interactions to reproduce a competitor’s capabilities, demonstrates the growing challenge of
  • Model distillation,
  • Capability extraction and
  • Fraudulent API use.
• AI-enabled surveillance creates another concern. As AI becomes more powerful in analysing and coordinating information,
  • It could facilitate excessive control by powerful actors,
  • Raising questions concerning the Right to Privacy under Article 21 of the Indian Constitution.
• These developments demonstrate that AI governance must address not only the content generated by AI but also the infrastructure, access, autonomy and downstream actions enabled by AI systems.

Proposed Guardrails for India:

• Mandatory misuse reporting: AI platforms above a specified scale should be legally required to report detected misuse to CERT-In and a designated AI Safety Authority.
• Watermarking: AI-generated content, particularly in political and public-interest contexts, should carry mandatory provenance or watermarking mechanisms to improve traceability.
• Regulation of Agentic AI: Existing platform-accountability frameworks must explicitly cover agentic AI—systems capable of taking actions in the real world rather than merely generating information.
• Criminalising systematic distillation: New legislation should prohibit and criminalise systematic unauthorised model distillation and fraudulent mass API access.
• Statutory AI safety institution: Voluntary disclosure should be replaced or supplemented by an independent statutory authority empowered to -
  • Compel disclosure;
  • Conduct audits;
  • Investigate systemic risks; and
  • Impose appropriate restrictions.

India’s Opportunity:

• India can play a significant role in developing a governance model that combines innovation with accountability.
• Its experience with Digital Public Infrastructure (DPI), UPI, Aadhaar and platform regulation demonstrates the ability to govern technology at population scale.
• The objective should be to ensure that AI remains aligned with democratic institutions, individual rights, privacy, public safety and broad-based societal interests.
• This will prevent AI from becoming concentrated in the hands of a few powerful technology companies.

Conclusion:

• The emergence of agentic and increasingly autonomous AI means that governance frameworks designed only for conventional social-media platforms or generative content are becoming insufficient.
• India therefore needs a risk-based, legally enforceable and technology-neutral AI governance framework combining innovation, transparency, accountability, cybersecurity and fundamental rights.
• The central lesson is clear: AI development is accelerating, but governance must accelerate faster.