OpenAI will stop training its AI models on the internet

OpenAI-access-for-frontier-model-training-do0pch) has announced it will no longer train its artificial intelligence models using real-time data from the internet. This decision follows comments from one of its engineers indicating that such an approach would not be effective, but now concerns have arisen regarding the risks associated with this strategy.

The measure taken by OpenAI means that its models will not receive live web data, potentially limiting the freshness of the information they process. However, the company argues that this disconnection is necessary to mitigate significant risks related to the quality and reliability of the data used to train its AI systems.

## Why it matters This decision marks an important shift in how OpenAI develops its models, which could impact the quality of the responses they provide. By not training with live data, OpenAI aims to avoid the inclusion of erroneous or biased information that could arise from the chaotic and unregulated nature of the internet.

## What we know OpenAI has indicated that its decision is based on the need to mitigate risks, although it has not specified what types of risks have been identified. The company seems to prioritize the integrity and safety of its models over the quantity of data they can process.

## What is still unclear It has not yet been confirmed how this strategy will affect the long-term evolution of OpenAI's models. It remains unknown whether this disconnection will influence the models' ability to adapt to new trends or updated information, which could be a key factor for their relevance in the future.