Enabling AI Safety Information Sharing: UK Competition Law Block Exemptions and Institutional Design

Jacob Schaal, Destiny Muller, Michael Chen, Liha Leung

Keywords: AI safety, information sharing, competition law, institutional design, antitrust

TL;DR: We classify AI safety information and design competition law block exemptions enabling labs to share critical information without antitrust violations.

Abstract:

Frontier AI labs face a coordination failure: sharing safety-critical information1
could prevent systemic failures, but competition law designed to prevent collusion2
creates legal barriers to collaboration. This paper addresses this coordination failure3
through comparative institutional analysis and legal framework redesign for the4
UK ecosystem. Drawing on cybersecurity (ISACs/ISAOs, CISA) and pharma-5
ceutical (EudraVigilance) precedents from the UK, EU, and US, we demonstrate6
how sector-specific legal exemptions paired with neutral clearinghouse institutions7
resolve tensions between competition enforcement and safety-critical information8
exchanges. We develop a two-dimensional taxonomy that maps technical AI infor-9
mation by commercial sensitivity and safety relevance, enabling clearinghouses10
and competition authorities to weigh antitrust risk against safety value. Applying11
UK Competition Act Chapter I analysis reveals that safety-critical information12
exchanges currently lack legal clarity; most crucially, the existing R&D Block13
Exemption Order (2022) does not protect post-deployment disclosures. Our analy-14
sis demonstrates that effective block exemptions require three design principles:15
(1) FRAND access, (2) anonymisation through neutral intermediaries, and (3)16
transparency requirements. We propose establishing the UK AI Security Institute17
(AISI) as a neutral clearinghouse and systematically evaluate nine institutional18
mechanisms to incentivise AI lab information sharing.

Submission Number: 26