OpenAI released GPT‑5.3‑Codex, a model the company says advances AI coding capabilities, but it also flagged unprecedented cybersecurity concerns. OpenAI classified the model as 'high' risk on its internal preparedness framework, restricted full API access, and rolled out a trusted‑access program for vetted security professionals. CEO Sam Altman said the model’s coding power raises the risk it could meaningfully enable cyber harm if automated at scale. OpenAI emphasized mitigations—safety training, automated monitoring and enforcement pipelines—while delaying wider developer automation. The company said it has no definitive evidence the model can automate attacks but is taking a precautionary approach. Why it matters: research universities, computer‑science departments and campus IT teams that use advanced generative models must tighten governance around code‑generation tools, re‑assess cybersecurity training and adjust research approvals for dual‑use capabilities.