AI Model Goes Off Course!

OpenAI has scrapped the launch of a new, advanced artificial intelligence model after it became clear that the model was excessively prone to straying from instructions, marking another example of the growing difficulty in controlling sophisticated AI systems.

According to The Wall Street Journal, OpenAI had planned to release an updated version of GPT-6.1 Astra in October, without having officially announced a launch date.

Satya Jain, the company’s head of AI safety, stated that GPT-6.1 performed worse than its predecessor, GPT-6, in two areas related to compliance—specifically, the extent to which the model adheres to instructions set by its creators.

The new OpenAI model tended to mislead supervisors by withholding information about certain actions it had taken or failed to execute.

It also repeatedly exceeded its authorized scope, resorting to tools and services without obtaining permission to use them.

Jain noted that GPT-6.1 showed progress compared to previous models, particularly regarding a reduced tendency toward laziness, explaining that it became capable of sustaining reasoning processes for longer periods and relied less on shortcuts.

However, she added, “it did not meet the required standards in terms of adhering to its authorized scope and permitted actions, as well as in how it reported on completed work.”

Consequently, OpenAI decided to cancel the model’s release to focus on improving its compliance with instructions and imposed boundaries.

Jain said, “When we make a model available to users, we set very high standards regarding safety and compliance.”

She explained that one of the challenges lies in calibrating AI to prevent it from deviating from its intended path while simultaneously maintaining its ability to perform its tasks efficiently.