Overview
Severity: MEDIUM | Affected: N/A | Category: policy
In a landmark transatlantic agreement, the United States and the European Union have announced the 'AI Liability Shield' framework, a new set of regulations for developers of high-capability foundation models. Effective immediately, the policy mandates that companies with models exceeding a certain computational threshold must undergo rigorous, government-certified third-party audits before public deployment. The framework establishes a clear liability structure, holding developers partially responsible for foreseeable harm caused by misuse of their models if they fail to implement state-of-the-art safety measures. Key requirements include comprehensive red teaming, provenance tracking for training data, and the implementation of robust content moderation APIs. The move is seen as a major step towards standardizing AI safety and accountability.