Anthropic CEO warns a rogue AI 'swarm' could seize the internet within a year and cause 'hundreds of billions' in damage

Roughly 700 OpenAI test agents built their own message board, called themselves a collective and hacked a major AI platform.