OpenAI admits new models likely to pose 'high' cybersecurity risk

OpenAI admits new models likely to pose 'high' cybersecurity risk
Better models also mean higher risk, but there are mitigations.