Trump: Safety rule for AI is sufficient

Donald Trump, former president of the United States, recently claimed in a speech that he is the only safety rule that artificial intelligence (AI) needs. This statement coincides with his opposition to the introduction of extra regulations for algorithms in the US, which stands in stark contrast to the approach of the European AI regulation. This European legislation imposes strict obligations on governments and companies working with AI systems, such as procurement and oversight, and forces them to comply with specific safety standards.

In his approach to AI safety, Trump rejects all formally established regulations and presents his own leadership as a sufficient guarantee for the safe use of AI. However, this is more political rhetoric than a developed policy proposal; there are no guidelines on enforcement, timelines, or checks. Safety around AI encompasses essential elements such as risk assessment and transparency of data models, where currently in the US only voluntary guidelines like the NIST AI Risk Management Framework exist.

In contrast to the American approach, the European AI Act opts for binding rules based on the risk level of AI applications. High-risk applications in sectors such as healthcare and government are faced with strict requirements like documentation and human oversight, meaning that organizations using American models like OpenAI’s GPT-4 or Google Gemini must comply with these EU rules. It is therefore concerning that Trump's deregulatory proposal could lead to a larger gap between regulations in the US and Europe, creating additional compliance challenges and costs for European companies.

For tech companies operating in both the US and the EU, there is a balance between different regulations. American tech companies like OpenAI and Meta are already implementing safety measures, but without clear minimum standards from the US, they are responsible for their own policies. This has direct implications for Dutch companies and government institutions, which must adhere to the higher demands set by the EU, including risk management and policy documentation. In the coming years, it will become evident whether the gap between American and European AI regulations will narrow or widen.

Read the full article from AI Insider.