[Paper Review] Dual Governance: The intersection of centralized regulation and crowdsourced safety mechanisms for Generative AI

Avijit Ghosh, Dhanya Lakshmi|arXiv (Cornell University)|Aug 2, 2023

Law, AI, and Intellectual PropertyComputer Science3 citations

TL;DR

This paper proposes Dual Governance, a hybrid framework integrating centralized U.S. regulations with community-driven, open-source safety mechanisms for generative AI. By linking regulatory standards to technical tools via a transparent registry, it enhances clarity, uniformity, availability, and nimbleness—enabling equitable innovation while mitigating ethical harms like misinformation, bias, and intellectual property violations.

ABSTRACT

Generative Artificial Intelligence (AI) has seen mainstream adoption lately, especially in the form of consumer-facing, open-ended, text and image generating models. However, the use of such systems raises significant ethical and safety concerns, including privacy violations, misinformation and intellectual property theft. The potential for generative AI to displace human creativity and livelihoods has also been under intense scrutiny. To mitigate these risks, there is an urgent need of policies and regulations responsible and ethical development in the field of generative AI. Existing and proposed centralized regulations by governments to rein in AI face criticisms such as not having sufficient clarity or uniformity, lack of interoperability across lines of jurisdictions, restricting innovation, and hindering free market competition. Decentralized protections via crowdsourced safety tools and mechanisms are a potential alternative. However, they have clear deficiencies in terms of lack of adequacy of oversight and difficulty of enforcement of ethical and safety standards, and are thus not enough by themselves as a regulation mechanism. We propose a marriage of these two strategies via a framework we call Dual Governance. This framework proposes a cooperative synergy between centralized government regulations in a U.S. specific context and safety mechanisms developed by the community to protect stakeholders from the harms of generative AI. By implementing the Dual Governance framework, we posit that innovation and creativity can be promoted while ensuring safe and ethical deployment of generative AI.

Motivation & Objective

• Address the limitations of purely centralized AI regulations, which often lack technical specificity, uniformity, and agility.
• Overcome the shortcomings of decentralized, crowdsourced safety tools, which lack enforcement, oversight, and standardization.
• Create a synergistic governance model that combines regulatory authority with community-driven technical interventions for generative AI safety.
• Ensure equitable access to safety mechanisms for both large and small technology developers and end-users.
• Establish a dynamic, reviewable system that adapts to evolving AI technologies and regulatory needs.

Proposed method

• Develop a centralized registry that maps U.S. regulatory requirements to specific, open-source safety tools (e.g., watermarking, model erasure, dataset filtering).
• Integrate existing open-source tools—such as glaze for model evasion, erasure for concept removal, and LLM watermarking—into a standardized governance framework.
• Use periodic policy and tool reviews to maintain relevance and adapt to technological and regulatory changes.
• Incorporate whistleblower and human alternative mechanisms inspired by the Blueprint for an AI Bill of Rights to ensure accountability.
• Establish a transparent, auditable system where regulatory intent aligns with technical implementation through documented mappings.
• Leverage existing regulatory bodies (e.g., CFPB) to enable actionable recourse and enforce compliance with safety standards.

Experimental results

Research questions

• RQ1 How can centralized regulations and decentralized safety mechanisms be meaningfully integrated to improve the governance of generative AI?
• RQ2 What criteria are necessary for a governance framework to ensure clarity, uniformity, availability, nimbleness, and transparency in AI safety?
• RQ3 To what extent can open-source safety tools be effectively aligned with government regulations to create a cohesive regulatory ecosystem?
• RQ4 How can the framework maintain relevance and adaptability in the face of rapid technological change in generative AI?
• RQ5 What mechanisms ensure that both large corporations and smaller entities can equitably access and implement safety measures?

Key findings

• The Dual Governance framework successfully addresses key shortcomings of both centralized regulation and crowdsourced safety by combining their strengths into a unified, transparent system.
• Mapping regulations to specific technical tools enhances clarity and ensures that regulatory intent is technically implementable and verifiable.
• Periodic reviews of both regulations and tools enable the framework to remain agile and responsive to emerging AI risks and innovations.
• The framework supports equitable innovation by making safety mechanisms accessible and cost-effective for small and medium-sized developers.
• By incorporating mechanisms for actionable recourse and transparency, the framework improves accountability and user trust in generative AI systems.
• The model demonstrates that a hybrid governance approach can achieve higher levels of uniformity, availability, and nimbleness than either approach alone.