OpenAI, the creator of ChatGPT, has overhauled its safety protocols and halted numerous training runs due to concerns about the cyber capabilities of its upcoming Astra model. The company believes Astra may have reached a “critical” level of cyber proficiency, prompting the precautionary measures.
OpenAI has not publicly detailed the specific cyber capabilities that triggered the pause, but the decision indicates a significant internal concern about the potential risks associated with increasingly powerful AI agents. The company is tightening its internal safeguards to address these risks before resuming full-scale training.
The pause affects a “significant number” of training runs, suggesting a broad reassessment of safety measures is underway. This move highlights the challenges developers face in controlling and predicting the behavior of advanced AI systems as they approach increasingly sophisticated levels of functionality. OpenAI’s actions demonstrate a proactive approach to mitigating potential harms before the Astra model is released.
We don't rate truth. We strip the spin and show you which perspectives covered the story. You decide.
Read the original coverage
💬 Comments
📜 Comment Policy