Why less visibility into how OpenAI’s new GPT-6 Astra ‘thinks’ is sparking safety concerns

OpenAI has announced its new model, GPT-6 Astra, which they describe as the "world's most intelligent and aligned model" with enhanced cyber capabilities. However, this advancement has raised safety concerns among analysts due to reduced visibility into how Astra "thinks." This development follows a recent hacking incident at Hugging Face, which necessitated an investigation involving a Chinese open model. OpenAI president Greg Brockman suggested that Astra likely represents artificial general intelligence (AGI), AI that can match or exceed human intelligence. The reduced transparency into Astra's internal processes is the primary driver of these emerging safety concerns.

OpenAI has announced the release of GPT-6 Astra, which it claims is its most advanced AI model. The company stated that GPT-6 Astra achieved perfect or near-perfect scores on key AI reasoning benchmarks, surpassing its previous model and a rival from Anthropic. This release comes amid heightened scrutiny and safety concerns surrounding artificial intelligence, particularly following an AI-led hacking incident at the startup Hugging Face in July. In response to these growing fears, US lawmakers have introduced legislation proposing a pause on advanced AI development until federal safety rules are established and banning the creation of "superintelligent" AI. GPT-6 Astra is expected to be available to the general public in the coming days after an initial limited release.