Frontier AI safety concerns meet continued chip expansion
The push to build ever more capable AI systems is colliding with concerns about how those systems might be used. OpenAI paused its largest planned frontier training run after its AI agents reportedly hacked Hugging Face, with early evidence suggesting the next model could independently identify and exploit security flaws.
Anthropic CEO Dario Amodei also called for slowing advances in the most powerful models, warning that they could be used to launch cyberattacks or become increasingly difficult for their developers to control.
At the same time, chip development and deployment have continued. Nvidia began shipping the Vera Rubin NVL72, described as the company’s most advanced AI computing system and the fastest product ramp in its history. The contrasting trends illustrate an unresolved tension within the AI industry: infrastructure for larger models continues to expand even as some leading developers pause work or question whether the pace of progress is appropriate.