Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Securing the Future of GenAI: Policy and Technology

Mihai Christodorescu, Ryan Craven, Soheil Feizi, Neil Gong, Mia Hoffmann, Somesh Jha, Zhengyuan Jiang, Mehrdad Saberi Kamarposhti, John Mitchell, Jessica Newman, Emelia Probasco, Yanjun Qi, Khawaja Shams, Matthew Turek

TL;DR

This paper analyzes the policy landscape for GenAI across the EU, PRC, and US, and situates it within multilateral governance and military risk-management insights. It synthesizes workshop discussions on three core risk-mitigation pillars—alignment, provenance/watermarking, and interpretability—while candidly acknowledging their technical limitations and gaps relative to policy goals. The authors advocate a risk-based, interdisciplinary, and iterative approach to governance, emphasizing agile regulation, failure-sharing, and the development of out-of-model safety guardrails. The work highlights the urgency of aligning rapid GenAI development with safeguard mechanisms to enable safe innovation across diverse jurisdictions and applications.

Abstract

The rise of Generative AI (GenAI) brings about transformative potential across sectors, but its dual-use nature also amplifies risks. Governments globally are grappling with the challenge of regulating GenAI, balancing innovation against safety. China, the United States (US), and the European Union (EU) are at the forefront with initiatives like the Management of Algorithmic Recommendations, the Executive Order, and the AI Act, respectively. However, the rapid evolution of GenAI capabilities often outpaces the development of comprehensive safety measures, creating a gap between regulatory needs and technical advancements. A workshop co-organized by Google, University of Wisconsin, Madison (UW-Madison), and Stanford University aimed to bridge this gap between GenAI policy and technology. The diverse stakeholders of the GenAI space -- from the public and governments to academia and industry -- make any safety measures under consideration more complex, as both technical feasibility and regulatory guidance must be realized. This paper summarizes the discussions during the workshop which addressed questions, such as: How regulation can be designed without hindering technological progress? How technology can evolve to meet regulatory standards? The interplay between legislation and technology is a very vast topic, and we don't claim that this paper is a comprehensive treatment on this topic. This paper is meant to capture findings based on the workshop, and hopefully, can guide discussion on this topic.

Securing the Future of GenAI: Policy and Technology

TL;DR

This paper analyzes the policy landscape for GenAI across the EU, PRC, and US, and situates it within multilateral governance and military risk-management insights. It synthesizes workshop discussions on three core risk-mitigation pillars—alignment, provenance/watermarking, and interpretability—while candidly acknowledging their technical limitations and gaps relative to policy goals. The authors advocate a risk-based, interdisciplinary, and iterative approach to governance, emphasizing agile regulation, failure-sharing, and the development of out-of-model safety guardrails. The work highlights the urgency of aligning rapid GenAI development with safeguard mechanisms to enable safe innovation across diverse jurisdictions and applications.

Abstract

The rise of Generative AI (GenAI) brings about transformative potential across sectors, but its dual-use nature also amplifies risks. Governments globally are grappling with the challenge of regulating GenAI, balancing innovation against safety. China, the United States (US), and the European Union (EU) are at the forefront with initiatives like the Management of Algorithmic Recommendations, the Executive Order, and the AI Act, respectively. However, the rapid evolution of GenAI capabilities often outpaces the development of comprehensive safety measures, creating a gap between regulatory needs and technical advancements. A workshop co-organized by Google, University of Wisconsin, Madison (UW-Madison), and Stanford University aimed to bridge this gap between GenAI policy and technology. The diverse stakeholders of the GenAI space -- from the public and governments to academia and industry -- make any safety measures under consideration more complex, as both technical feasibility and regulatory guidance must be realized. This paper summarizes the discussions during the workshop which addressed questions, such as: How regulation can be designed without hindering technological progress? How technology can evolve to meet regulatory standards? The interplay between legislation and technology is a very vast topic, and we don't claim that this paper is a comprehensive treatment on this topic. This paper is meant to capture findings based on the workshop, and hopefully, can guide discussion on this topic.
Paper Structure (32 sections, 4 figures, 1 table)

This paper contains 32 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Regulatory-Policy Considerations for GenAI
  3. The Policy Landscape
  4. Scope.
  5. The PRC’s Interim Measures for the Management of GenAI.
  6. The European Union’s AI Act.
  7. The United States Executive Order on AI.
  8. Divergent Priorities.
  9. Multilateral Governance
  10. Learning from Military Risk Management
  11. Risk Mitigation through Model Alignment
  12. Challenges in Achieving Alignment
  13. Risk Mitigation through Model Inspection
  14. Use Cases.
  15. Inner Interpretability and Outer Explainability.
  16. ...and 17 more sections

Figures (4)

  • Figure 1: The software stack of GenAI-powered systems (shown here simplified to focus only on the components that can directly impact GenAI security) can have a variety of stakeholders, depending on distribution model. Data and compute providers have different leverage towards ensuring the security and safety of GenAI, compared to model providers and to app builders. Examples of GenAI apps were based on https://www.sequoiacap.com/article/generative-ai-act-two/ .
  • Figure 2: Deepfakes can be used to promote investment scams. This screenshot is from a deepfake video that circulated in November 2023 on social media, primarily targetting South African users, in which Bongiwe Zwane and Francis Herd from the South African Broadcasting Corporation (SABC, South Africa’s public TV and radio broadcaster) and Elon Musk appeared to promote an investment opportunity. The video is staged as a news clip, with a brief introduction from (deepfake) Herd, followed by (deepfake) Musk announcing "powerful, world-first investment software" while on a stage. SABC, Zwane, and Herd all denounced the deepfake through their web and social media presence (see https://www.sabcnews.com/sabcnews/894115-2/, https://www.facebook.com/bongiwe.khumalo.946/videos/1151663159551664, and https://twitter.com/FrancisHerd/status/1721835389994799321). Screenshot and details from https://africacheck.org/fact-checks/meta-programme-fact-checks/beware-another-elon-musk-investment-scam-using-deepfake .
  • Figure 3: Media manipulation has been happening long before GenAI. In the above example from 1902, three Civil War photos were combined to create a fake image of General Ulysses S. Grant. Images from Library of Congress, information from https://www.npr.org/sections/npr-history-dept/2015/10/27/452089384/a-very-weird-photo-of-ulysses-s-grant .
  • Figure 4: Notional adversarial landscape for media falsification. The x-axis shows various adversaries increasing in capability. The y-axis shows resources available to these classes of adversaries.