OpenAI calls for a delay in AI development

OpenAI has announced that it will delay the development of its most advanced AI systems until stricter safety standards and independent tests are established. This decision is important in the context of the new European AI regulation, which imposes additional obligations on both providers and users of AI technologies. The aim is to mitigate risks such as accidents, misuse, and misinformation.

The company, known for its products like ChatGPT and the multimodal GPT-4o, wants to ensure that the scaling of its systems is linked to meeting safety criteria. This primarily concerns the so-called frontier models, which can have a significant societal impact. OpenAI advocates for common safety standards and external audits conducted by independent third parties, so that the models can be tested against clearly defined criteria.

The European AI Act, which categorizes systems based on risk classes, requires additional documentation and audits for generative models with systemic risks. This means that OpenAI must provide technical documentation and a summary of the training data for European applications. For governments, healthcare providers, and schools using AI technologies, this means they can rely on structured documentation and processes, facilitating compliance and procurement.

The challenging balance between preventing rapidly growing models without adequate safety nets and ensuring a competitive market is emphasized. The delay in development may lead to a slower rollout of new features, but also to reduced system risks. In a competitive field, where players like Google and Meta are also active, this could ultimately ensure improved safety and usability of AI applications in public services.

Read the full article from AI Insider.