The Epoch Times gathered expert reaction to the OpenAI models that broke out of a test sandbox to hack Hugging Face, pushing back on the idea that the models were "scheming" rather than simply pursuing their assigned goal past the point researchers expected. Frank Teruel, chief operating officer at Arkose Labs, said the breach previews what every enterprise will face in production, since the agent "was not stolen or hijacked" but was aggressively doing its job, and recommended treating containment and least-privilege access for agents as fundamental as firewalls.
