Annual discrimination impact assessments

Notify individuals that automated systems are being used

Right to opt out of solely automated decisions and appeal

AG enforcement and private right of action

Insurer must demonstrate algorithm does not unfairly discriminate

Maintain governance framework for algorithmic fairness in insurance

Employers and AI vendors share liability for discriminatory outcomes

AI tools must not screen out individuals with disabilities (ADA)

Risk management system with bias testing throughout the AI lifecycle

Training data must be examined for biases and be representative

Comprehensive technical documentation

Human oversight mechanisms enabling review and override

Fundamental rights impact assessment (FRIA)

Register high-risk AI systems in EU-wide database

Ongoing monitoring for accuracy, robustness, and population drift

Fines up to EUR 35M or 7% of global annual turnover

Joint pay assessment triggered at 5% gender pay gap

Publish gender pay gap data and pay structure information

Mandatory reporting for frontier AI model developers

Adverse action notices must explain AI-based credit decisions

Algorithmic disgorgement — forced deletion of models trained on improperly obtained data

Right not to be subject to solely automated decisions with significant effects

Right to obtain human intervention and contest the decision

Fines up to EUR 20M or 4% of global annual turnover

Notify candidates and obtain consent before AI-analyzed video interviews

Delete video recordings within 14 days of candidate request

Written consent required for facial recognition in hiring

Annual independent bias audit before deployment

Publish audit results on employer website for 2+ years

Notify candidates that an AEDT will be used

Up to $1,500 per violation per day

Four-Fifths (80%) Rule for adverse impact

Statistical significance corroborates adverse impact findings

Employer must demonstrate business necessity if adverse impact found