EU AI Act
A risk- and role-based regulatory framework with phased obligations. The high-risk classification, applicable provisions and your role must be determined for each AI system.
Find which requirements may affect your AI systems, see the evidence reviewers need, and understand how HAIEC can help establish it.
Explore official sources, important milestones and technical evidence paths. Move from a regulatory question to a clearer assurance plan without mistaking a mapping for a compliance verdict.
Inspect a reference to see what its requirements may involve, which evidence HAIEC can contribute within scope, and what remains an organizational or independent-review responsibility.
A risk- and role-based regulatory framework with phased obligations. The high-risk classification, applicable provisions and your role must be determined for each AI system.
Roles identified by the reference: Provider, deployer, importer and other roles depending on the obligation. Your selected perspective is an AI deploying or using team; confirm the legally defined role separately.
Annex III: 2 Dec 2027; Annex I: 2 Aug 2028
Role under the Act, system type and risk classification, territorial scope and relevant high-risk use.
Selected official-source milestones, not a continuously updated legal feed. Distinguish enacted future dates from an effective law and from proposals.
Milestones reflect the referenced instruments and selected provisions. The date a requirement applies to a specific organization or system needs separate review. EU official timeline ↗
Explore enacted laws, a sector-led regulatory landscape and a historical proposal without treating them as equivalent legal obligations.
A risk- and role-based regulatory framework with phased obligations. The high-risk classification, applicable provisions and your role must be determined for each AI system.
Data protection requirements can apply when AI processing involves personal data within the GDPR’s territorial and material scope.
Colorado’s SB 26-189 enacted an updated framework for covered automated decision-making technologies used in consequential decisions involving Colorado consumers.
California’s final CCPA regulations include specific ADMT obligations for businesses using covered technologies to make significant decisions about consumers.
NYC regulates use of covered automated employment decision tools for screening candidates or employees in employment decisions.
California privacy rules can apply to businesses meeting specified thresholds that handle California consumers’ personal information.
HIPAA privacy and security obligations can apply to covered entities and business associates handling protected health information.
The UK uses existing laws and sector regulators rather than one generally applicable EU-style AI Act.
The proposed Artificial Intelligence and Data Act in Bill C-27 did not become law. Existing Canadian privacy and sector-specific laws may nevertheless apply to AI activities.
HAIEC examines consequential AI through independent evidence views. A regulatory obligation is not automatically satisfied by scanning code, running a test or issuing a receipt.
What can your AI request, what is authorized, what permissions exist, what can the code enable and what was actually observed?
These are independent evidence views, not five mandatory sequential steps.
Registered AI Systems, repositories and supported agent-to-tool and resource relationships.
Scoped static source findings and authorized runtime test responses, with coverage and producer identity.
Delegation context, effective grants, control findings and human decision records where supported and connected.
Qualified Assurance evaluations, reports and receipts with stated scope and limitations.
Present evidence for human, legal, auditor or certification-body review without substituting for those decisions.
Standards, professional attestation criteria, voluntary guidance and HAIEC-authored methodology are useful references. They are not interchangeable with enacted regulations.
Risk management guidance for governing, mapping, measuring and managing AI risks.
View framework →AI management-system requirements. Accredited third-party certification, where sought, is separate from HAIEC evaluation.
Official ISO source ↗AICPA Trust Services Criteria for independent CPA examinations. Technical evidence may contribute but does not create a SOC 2 report.
View SOC 2 guidance →Security risk taxonomies and implementation guidance, not a legal compliance or certification program.
OWASP source ↗Information-security management-system requirements, separate from an AI-specific regulatory determination.
View framework →HAIEC's evidence and assurance methodology. It is not government regulation or a third-party accreditation.
Explore methodology →Start with the AI system and the decision you need to support. HAIEC helps you connect technical findings, authority context and qualified evidence to a defensible assurance workflow.