OpenAI's Atlas shrugs off security concerns over prompt injection
'Trust no AI' says one researcher OpenAI's brand new Atlas browser is more than willing to follow commands maliciously embedded in a web page, an attack type known as indirect prompt injection....