Building Governance-Support Systems for Global AI Governance: TCA's Proposal for FlexHEG
Abstract: Last Friday, we submitted a grant proposal to The Mechanisms for Flexible Hardware-Enabled Guarantees (FlexHEG) call from Jaan Tallinn’s Survival and Flourish Foundation. FlexHEG aims to develop critical hardware-based systems for verifying and enforcing compliance of advanced AI systems with safety and ethical guidelines. Our submitted 4-page proposal, The Trustless Computing Certification Body and Seevik Controls Project” - offers a comprehensive approach to this challenge based on our Trustless Computing Paradigms.
Leveraging nine years of work almost exclusively focuses on such challenge, and a wide global network of experts and R&D partners, we propose to create high-level paradigms, architectures, and governance mechanisms to implement FlexHEG and support other funded project that will work on technical specifications, prototype and validation of technical and socio-technical components.
As we stand on the precipice of an AI revolution, the risks to human safety and the potential for an unprecedented concentration of power loom larger than ever before. The rapid advancement of artificial intelligence technologies presents both immense opportunities and existential threats to humanity. These risks are not just theoretical – they are imminent and demand immediate action.
The scale and nature of these challenges are such that only a strong, comprehensive global treaty can adequately address them. Much like the challenges posed by nuclear weapons in the mid-20th century, AI requires a coordinated international response. However, the complexity and rapid evolution of AI technologies make this an even more daunting task than nuclear disarmament or climate change mitigation.
To achieve an effective global governance framework for AI, three critical elements are required:
Broad Consensus: We need a wider agreement among states and scientists on the nature and immediacy of AI risks. This includes acknowledging both the potential for catastrophic accidents and the dangers of unaccountable concentration of power.
Trustworthy Mechanisms: There must be consensus that socio-technical systems and governance mechanisms can be developed that are both trustworthy and widely trusted across geopolitical divides.
Effective Treaty-Making: We require a treaty-making process that is far more effective, comprehensive, and timely than what we've seen for nuclear non-proliferation or climate change. The rapid pace of AI development demands unprecedented speed and depth in international negotiations and agreement implementation.
Trustworthy Mechanisms
The FlexHEG initiative represents a crucial step towards ensuring the safety and accountability of advanced AI systems. It aims to develop hardware-enabled mechanisms for verifying and enforcing compliance of AI systems with agreed-upon rules and safety measures. At TCA, we recognize that this goal is not just important – it is essential for the future of humanity.
Our approach to the FlexHEG challenge is unique and comprehensive, building on nearly a decade of pioneering work in ultra-secure computing and global IT governance. Here's what sets our proposal apart:
Unparalleled Expertise: Since 2015, TCA has been at the forefront of developing the Trustless Computing Paradigms (TCP) - a set of principles designed to achieve unprecedented levels of security and trustworthiness in critical IT systems. This experience uniquely positions us to tackle the complex challenges posed by FlexHEG. Our deep understanding of high-assurance hardware-based security, coupled with our insights into AI governance, provides a solid foundation for developing effective FlexHEG mechanisms.
Holistic Approach: Our proposal goes beyond mere technical solutions. We recognize that true security and trustworthiness stem from a combination of technical, socio-technical, and governance measures. Our approach integrates hardware security, democratic oversight, and innovative governance mechanisms. This comprehensive strategy ensures that the resulting FlexHEG systems will be not only technically sound but also socially acceptable and politically viable.
Global Perspective: Through our Free and Safe in Cyberspace conference series and extensive network of international partners, we bring a truly global perspective to this challenge. Our proposal aims to create solutions that can be trusted and adopted across geopolitical divides. This international outlook is crucial for developing FlexHEG mechanisms that can be universally implemented and trusted.
Synergy with Existing Initiatives: The FlexHEG proposal is not a standalone effort, but part of a broader vision for global AI governance and secure communications. It aligns closely with our Trustless Computing Certification Body (TCCB) initiative and the Harnessing AI Risk Initiative (HARI), now advanced primarily through the Coalition for a Baruch Plan for AI (CBPAI). This integration ensures that our FlexHEG work is grounded in a comprehensive framework for AI safety and global digital security.
Innovative Governance: We propose novel mechanisms for international oversight, including citizen-witness and citizen-jury processes. These approaches aim to foster deep trust among states and citizens - a crucial factor for the wide adoption of the strict controls envisioned by FlexHEG. By involving diverse stakeholders in the governance process, we can create a system that is both effective and widely accepted.
Our proposal for FlexHEG represents a natural evolution of TCA's long-standing mission and the sought fruition of 9 year of work. The Trustless Computing Certification Body, which forms a core part of our proposal, was envisioned as one of three key agencies in our broader plan for global AI governance. By developing standards and certifications for ultra-secure IT systems, we aim to create a foundation for both AI control systems and secure diplomatic communications.
The synergy between our FlexHEG proposal, TCCB, HARI, and CBPAI creates a comprehensive approach to the challenges of AI safety and global digital security. While CBPAI focuses on the broader geopolitical and diplomatic aspects of AI governance, our FlexHEG work addresses the critical technical and socio-technical underpinnings necessary for any effective governance system.