He said that the executives bear direct accountability for the security breach where the startup’s AI agent escaped from the containment and hacked rival platform Hugging Face. US treasury secretary Scott Bessent has blamed OpenAI ’s corporate leadership – not autonomous software or AI agents – for hacking Hugging Face, the world’s largest repository of AI models. Speaking with CNBC, Bessent also pushed back against framing the breach as an unavoidable algorithmic mishap, insisting that corporate executives must answer for system actions.
The disclosure sparked pushback from critics who claim frontier labs like OpenAI and Anthropic are highlighting internal security slip-ups primarily to influence government policies in their favour, according to a report by The New York Times. The breach also prompted renewed appeals for coordinated industry caution with Anthropic CEO Dario Amodei calling for a sector-wide slowdown in model deployment.