Balancing innovation with safety is the central dilemma in regulating AI. Using the lens of women's digital safety, discuss how India can design a regulatory framework that is neither innovation-stifling nor toothless.

Digital harm is gendered by design: UN Women finds that 90–95% of deepfakes online are sexualised images of women, while nearly 1.8 billion women and girls lack legal protection against technology-facilitated abuse [1]. AI regulation must therefore be deterrent enough to protect dignity, yet light enough to sustain innovation.

Where India stands

• DPDP Act, 2023, operationalised by the DPDP Rules, 2025 (notified 14 November 2025), makes deepfake generation from a woman's images without consent unlawful processing [2].
• IT (Intermediary Guidelines) Amendment Rules, 2026 (in force 20 February 2026) tighten due-diligence duties, with safe harbour under Section 79 as the enforcement lever [3].
• India AI Governance Guidelines, 2026 create the AI Governance Group, Expert Committee and AI Safety Institute, favouring voluntary compliance plus techno-legal solutions [4].
• Institutional scaffolding: I4C, CERT-In, cybercrime.gov.in and 14,658 Women Help Desks [5].

Why it risks becoming toothless

• Voluntary norms carry no timelines or penalties for AI-specific harms [4].
• Grievance redressal is slow relative to the virality of non-consensual imagery; one delayed takedown is permanent damage.
• Low reporting due to stigma, jurisdictional ambiguity, and weak awareness beyond metros [5].
• The Data Protection Board is yet to be fully operational [2].

Designing a calibrated framework

• Risk-tiering: binding obligations only for high-risk uses — sexual deepfakes and non-consensual intimate imagery — leaving benign AI light-touch.
• Techno-legal by default: mandatory provenance watermarking and hash-matching to block re-uploads, shifting the burden from litigation to code [3].
• Conditional safe harbour: time-bound removal of flagged NCII plus standardised transparency reporting.
• Regulatory sandboxes under the AI Safety Institute, so startups test guardrails without compliance shock [4].
• Victim-centric capacity: sensitised Women Help Desks and single-window takedown [5].

A framework that regulates risk rather than technology protects women without penalising builders. Anchoring such calibration in Article 21's privacy guarantee, while sustaining India's AI ambition, converts the innovation–safety trade-off into a partnership.