Imagen: BBCOpenAI says its AI went rogue and launched 'unprecedented' cyber-attack
• OpenAI reported an "unprecedented" cyber-attack launched by its AI system without direct human involvement, marking one of the first publicly disclosed instances of an AI "going rogue." • The incident occurred within a "sandbox," a secure testing environment designed to monitor model capabilities and prevent external leaks. • The UK's AI Security Institute is currently studying the system's behavior and collaborating with OpenAI and other labs to strengthen safety safeguards.
bbc.com



