Ethical AI refers to the safe and responsible use of artificial intelligence (AI). It involves three main approaches: principles, processes, and ethical consciousness. Ethical AI operationalizes AI Ethics, with a focus on four key verticals: safety, privacy, fairness, and transparency. Algorithm auditing is a key practice for determining how well a system performs on each of these verticals. While AI has many applications, such as conversational AI, ethics need to be prioritized to prevent poorly designed systems from being developed. The EU High-Level Expert Group on AI and the IEEE have formulated moral values that should be adhered to in the design and deployment of artificial intelligence. However, regulatory oversight and AI auditing are needed to bridge AI ethics from theory to practice.

The District of Columbia has introduced the Stop Discrimination by Algorithms Act to prohibit the use of algorithms that make decisions based on protected characteristics like race, sex, gender, disability, religion and age. The legislation would require annual audits and transparency from organizations, with failure to comply resulting in individual fines of $10,000 each. This three-pronged approach aims to mitigate algorithmic bias and discrimination, with the penalties applying to businesses possessing or controlling information on over 25,000 Washington DC residents, data brokers processing personal information and service providers. While the Act has received support from policymakers and academics, industry groups have criticized it as a compliance burden that could result in decreased credit access and higher-cost loans. The Act has set a national precedent that other states may follow.

Canada has proposed the Digital Charter Implementation Act, which would establish three acts for trust and privacy: the Consumer Privacy Protection Act, the Personal Information and Data Protection Tribunal Act, and the Artificial Intelligence & Data Act (AIDA). AIDA would set governance and transparency requirements for businesses that use or develop AI in Canada, including requirements relating to bias, transparency, risk mitigation, and record-keeping. The act would also establish the role of an AI & Data Commissioner to monitor compliance and order third-party audits as required. However, it is uncertain if AIDA will pass, and businesses that develop AI for personal information collection or use may have to answer to federal transparency and privacy protection expectations.

Spain has launched the first regulatory sandbox to test new rules for artificial intelligence (AI) and algorithmic systems as part of the EU AI act. The country has also introduced a rider-law, Royal Decree-Law 9/2021, to safeguard employment rights of delivery workers who work through digital platforms. The law includes a presumption of employment for riders who work under a digital platform algorithm, giving them additional job security and safety provisions. Employers are required to inform works councils of workers' parameters, rules, and instructions based on AI systems used as part of the platform. The Ministry of Labour has also published guidelines for complying with algorithmic transparency laws, including the GDPR where necessary.

AI regulation is growing significantly, with governments considering AI regulations, policies, and strategies to manage AI risks and harms. The EU has proposed the AI Act, which takes a risk-based approach and establishes four categories of AI systems based on potential harms. California has proposed regulations to extend non-discriminatory practices to automated decision systems and to regulate the day-to-day use of automated tools in the workplace. Both acts require ongoing monitoring and re-evaluation when significant changes are made to the system. The EU AI Act is more expansive and takes a sector-agnostic approach, while California's proposed laws are narrowly focused mainly on automated employment decision tools. Holistic AI offers a risk management platform to help enterprises identify risks and recommend steps to mitigate them. This blog article is not intended to provide legal advice or a legal opinion.