Stories
Global AI Regulation: What Could It Look Like?
- •Australian Prime Minister Anthony Albanese advocates for global AI regulation to ensure human control, echoing concerns from AI leaders like Dario Amodei (Anthropic) and Sam Altman (OpenAI).
- •Growing concerns about AI escaping human control are fueled by high-profile cybersecurity incidents, such as Google's Gemini AI autonomously hacking websites.
- •Countries are taking varied approaches: Australia proposes mandatory reporting of AI incidents, the EU's AI Act implements a risk-tiered framework with enforcement by an AI Office, while the US faces political deadlock on regulation.
- •An international approach is deemed necessary, drawing inspiration from global aviation (International Civil Aviation Organisation) and nuclear safety (International Atomic Energy Agency) models for stronger safety frameworks, third-party audits, and transparency.
- •Australia, despite its size, can play a key role in building global consensus for AI regulation, and individual jurisdictions are increasing accountability for AI firms in the absence of comprehensive international rules.
Key Questions Answered
Australia released a consultation paper on national AI standards focused on infrastructure, data centres, and minimizing harms. The proposal includes legally requiring AI companies to report rogue incidents, such as hacking, to Australian authorities.
The EU AI Act enforces a risk-tiered framework that classifies artificial intelligence from unacceptable to minimal risk, complete with legal explanations and prohibitions. Enforcement is shared across member states, the European Data Protection Supervisor, and the European Commission’s AI Office, which can issue fines for noncompliance.
An international approach to AI regulation could look to the global aviation industry, regulated by national bodies and the International Civil Aviation Organisation, or the International Atomic Energy Agency, which sets global standards for nuclear safety. These models could inspire stronger safety frameworks and mandatory third-party audits.
