Three former OpenAI employees are warning that increasingly advanced artificial intelligence models could become harder to monitor, urging the company to strengthen safety oversight and work with independent auditors, according to a report.