As Artificial Intelligence becomes all-pervasive, it raises serious concerns about maintaining fairness, trust, data security, intellectual property rights, and human-centricity. A global survey of 350+ CEOs of large corporations found that 70 percent of those who said AI was a crucial technology in their organisation were most worried about protecting data privacy; their other concerns included cybersecurity, compliance, and environmental impact. 87 percent of CEOs emphasized the need for robust governance and risk management to ensure secure and responsible AI deployment.
Ethical AI is the answer. By embracing relevant practices, enterprises can not only mitigate the above-mentioned challenges but also demonstrate accountability, transparency, and trustworthiness, to stand out in an environment where maleficent actors are using deepfakes, misinformation, discrimination and fraud to harm society. This is why organizations should make ethical AI the cornerstone of corporate governance going forward.
An ethical AI framework for transparency and trust
Rigorous model testing and high-quality training data – clean, accurate, comprehensive, unbiased – are essential for producing fair and reliable outcomes. Organizations should work at making “AI black-boxes” more transparent and explainable so they can understand and elaborate on the algorithmic rationale, especially in high-involvement decisions such as medical diagnosis or loan approval. They may also consider collaborating with social scientists for insights into human behaviours and cultures to drive human-centric development.
A clear line of responsibility establishes accountability for AI outcomes, enabling the organization to respond quickly in an emergency, for example, when a system is breached, breaks down, or causes harm. There should be stringent controls to ensure AI systems do not expose sensitive or confidential data, and are compliant with all applicable data protection laws, such as GDPR, HIPAA, etc. Human oversight, especially of critical AI-powered decisions, is a necessary safeguard against negative consequences, that enables early intervention in times of need.
Not just compliance, but culture
Ethical AI leaders go beyond compliance, to adopt proactive AI governance. These organizations embed clearly defined, core ethical principles within their AI applications, which not only comply with regulatory mandates but also align with recommended standards. They follow a clear AI governance charter, describing objectives, scope, processes, etc., and a risk-based approach where controls are proportional to an application’s risk level.
Besides a clear accountability mechanism, ethical AI leaders have cross-functional AI governance councils drawn from streams, such as data science, IT, legal, risk management, and business. Ethical checks, documentation, and other elements of governance can be integrated within AI workflows right at the design stage, across phases of the AI lifecycle, from data sourcing and model development to testing, deployment, feedback, and monitoring. Repeatable, auditable procedures for bias testing and incident response can ensure that policies don’t remain on paper.
Above all, ethical AI leadership stems from an organization’s AI culture. Senior management should drive it from the top by demonstrating responsible AI use in day-to-day work. They should also nurture a culture where ethical AI is everyone’s responsibility. Finally, leaders need to make sure everyone in the organization is trained in ethical AI practices, such as using personal data only with consent, avoiding bias in prompts, declaring AI-generated content, etc.
Ethical exemplars
Ethical AI is linked to quantifiable improvement in business metrics: when companies adopt human-centered AI, they could improve customer and employee retention rates by 15-30 percent and 20-35 percent respectively; they could also have 40-60 percent lower regulatory compliance costs vis-à-vis reactive compliance scenarios. But far more important is the impact of ethical AI on corporate reputation and trust. Ethical AI leaders are trusted by customers, employees, and regulators alike, and are valued by their communities. They are differentiated not only by their superior business performance, but also by values such as integrity and human-centricity.
Authored by Sunil Kumar Dhareshwar, Executive Vice President, Infosys