Researchers reveal OpenAI agents uploaded malicious code to RubyGems in May, marking another rogue incident for autonomous AI ...