Companies operating across state borders face different AI compliance requirements in each jurisdiction. No formal harmonisation mechanism exists.
Voluntary. Core reference adopted by all states. Covers human rights, fairness, transparency, accountability, reliability, privacy, contestability, oversight.
Harmonisation across government levels. Not binding on private sector. Guides risk assessment alignment.
10 guardrails for AI development. Voluntary but signals future regulatory direction for high-risk AI.
Mandatory guardrails dropped after pushback. Relies on existing laws, voluntary frameworks, AI Safety Institute ($29.9M). CAIO deadline: July 2026.
5 governance failures shaped every current framework: no legal basis, removed human oversight, flawed algorithm, suppressed concerns, no accountability.
Article 2 high-risk obligations enforceable on AU companies serving EU markets. Penalties up to EUR 35M / 7% global turnover.