Ethical Concerns Raised Over AI in Public Services
A discussion on social media highlights growing public concern over the ethical implications of using AI chatbots in government services. One user questioned what safeguards are in place to prevent bias, discrimination, and privacy violations, reflecting a demand for greater transparency and algorithmic auditing.
- The European Union's AI Act, which entered into force in August 2024, establishes a risk-based framework for AI systems in both the public and private sectors. Systems deemed "high-risk," such as those used in law enforcement or for accessing public benefits, are subject to stringent requirements, while certain "unacceptable risk" applications like social scoring are prohibited entirely. - In the UK, the government's "Data and AI Ethics Framework" guides the public sector in the responsible use of AI. This framework was updated in December 2025 to expand its scope to cover AI and algorithmic technologies, adding principles like privacy, sustainability, and safety. - Estonia has become a leader in the governmental use of AI, with over 130 projects implemented in the public sector since 2019. Use cases include AI-driven tax fraud detection, a network of chatbots for citizen services called "Bürokratt," and automatic transcriptions of court hearings. - Algorithmic audits are emerging as a key mechanism to ensure fairness and accountability in public sector AI. These audits involve independent assessments of algorithmic systems to check for biases and ensure they align with ethical and technical standards. - A 2019 study by the National Institute of Standards and Technology (NIST) found that facial recognition systems, often used in law enforcement, were up to 100 times more likely to return a false positive for a non-white person than for a white person. - In the Netherlands, tax authorities using an AI system to detect fraud were found to have breached European data rules, leading to a major scandal. Similarly, in Australia, an automated system wrongly accused 400,000 people of welfare fraud. - Public sector digital services in the US are required to comply with accessibility standards like the Web Content Accessibility Guidelines (WCAG) 2.1. This mandate extends to AI-powered tools to ensure they are usable by people with disabilities. - Service design methodologies are being adapted to address the complexities of AI in public services, focusing on human-centered design to ensure AI tools are transparent, explainable, and accountable. This includes creating "trust-centric" AI personas and ensuring there is always a clear path to human support.
