OpenAI, Microsoft, Google: are companies losing control over AI?

In recent months, major technology companies such as OpenAI, Microsoft, and Google have faced various issues regarding their AI systems. Incidents with tools like Google Gemini and Microsoft Copilot have raised concerns about the control that companies have over these technologies. With the upcoming European AI regulation and the General Data Protection Regulation (GDPR), the urgency for companies to get their algorithms and data models in order is growing.

Recent examples, such as the errors with Google Gemini and the privacy issues with Microsoft Copilot, demonstrate that there are significant vulnerabilities in generative AI. This not only impacts user experience and trust but can also cause legal and reputational damage for the involved organizations. An incident where a chatbot from Air Canada provided incorrect information shows that a wrong answer from AI can be interpreted as a binding promise to consumers, underscoring the necessity for quality control and transparency.

Another issue is the lack of human oversight over AI systems. Many organizations rely on automated processes without sufficient supervision, resulting in unpredictable outcomes, especially in generative AI. The European AI regulation, which will be implemented in phases between 2025 and 2026, states that risk management and human oversight are necessary. This means that companies must maintain accurate documentation and data management to comply with the new requirements.

In the Netherlands, the Dutch Data Protection Authority monitors compliance with privacy standards in both the public and private sectors. This has led to the establishment of an algorithm oversight body and the publishing of algorithm registers by municipalities. The lessons from the Childcare Benefits Affair, such as the importance of transparency and human intervention, are more relevant than ever now that European legislation imposes detailed requirements on algorithms and AI. Companies and public institutions must now take action to improve their processes, technical solutions, and culture to ensure control over AI.

Read the full article from AI Insider.