OpenAI Accidentally Leaked Its Upcoming o1 Model to Anyone With a Certain Web Address
Briefly

As Tom's Guide reports, a number of users on X-formerly-Twitter discovered that a simple tweak to the URL allowed them to access the AI model. This unexpected access raised questions about how securely OpenAI manages its new releases and whether they are ready for broader public use, especially given the potential implications of AI models in various sectors.
The model was first announced in September and has only been available in 'preview' form to paying users since then. However, the leaked access demonstrates both the demand for advanced AI capabilities and the vulnerability that tech companies face in controlling access to their innovations.
Users were initially impressed by the purported model's capabilities, with one noting it could solve complex problems such as a mathematics challenge and even analyze an image of a SpaceX launch. Such capabilities suggest that OpenAI's o1 model could set new standards in AI functionality and performance.
Despite some impressive benchmarks, though, OpenAI recently found that its ability to provide correct - and not 'hallucinated' - answers still leaves plenty to be desired, indicating there are still significant challenges to overcome even with advanced models.
Read at Futurism
[
|
]