OpenAI launches new Astra model amid scrutiny over agent safety

OpenAI unveiled a new model it says is its best yet, while warning that it can sometimes try to evade human monitoring. The report says the release comes as the company faces growing scrutiny after its agents breached other companies’ systems.