In the absence of binding law, frontier AI firms are now drafting their own regulation. Critically examine whether industry-led AI regulation strengthens or weakens public accountability.

The U.S. has no federal AI law. To fill that gap, Anthropic is circulating a draft Bill among U.S. lawmakers. It would let Washington block a model's release or use if there is an imminent risk of killing more than 50 people [1]. Industry input can sharpen regulation. Accountability, however, depends on who finally decides, not on who writes the first draft.

How it strengthens accountability

• Fills a legal vacuum: a concrete text gives legislators something to debate where nothing existed [1].
• Technical expertise: developers understand what their own models can do, so rules can target real risks such as catastrophic misuse.
• Self-binding commitment: a firm accepting state power to block its own product breaks from the usual industry demand for no rules [1].
• Co-regulation precedent: India's AI Governance Guidelines also rely on industry-led voluntary frameworks, standards and audits [2].

How it weakens accountability

• Regulatory capture: the drafter decides where the line sits, even without any bad intent.
• Narrow trigger as a ceiling: a ">50 deaths" test leaves out widespread harms such as bias in lending or deepfake fraud, and may slow broader lawmaking [1].
• Unclear decider: the draft as reported does not say who judges whether a risk is "imminent" [1].
• Entry barriers: large firms can absorb compliance costs that start-ups cannot.
• No penalty for exit: voluntary codes bind only willing firms. India's graded liability is still only a recommendation [2].

Verdict Industry drafting strengthens accountability only when public institutions filter it. That means open legislative hearings, an independent public body holding the blocking power, and hearing critics who do not build AI. India's model rests on existing statutes (IT Rules 2021, DPDP Act 2023), with an AI Safety Institute and an AI Governance Group proposed [3]. The government states that the guidelines do not allow unrestricted deployment of high-risk AI [4].

Industry should be a co-author of AI rules, not the final authority. India should empower its AI Safety Institute to test frontier models, write graded liability into law and make incident reporting work [2]. Building on the Principal Scientific Adviser's techno-legal framework [5] would let "Innovation over Restraint" sit alongside public answerability. This upholds Article 21 and SDG 16's call for accountable institutions.