Anthropic tries to keep its new AI model away from cyberattackers as enterprises look to tame AI chaos

# Anthropic Keeps New AI Model Under Wraps Over Security Concerns Anthropic built an AI called Claude Mythos that's so skilled at finding computer security holes that the company decided not to release it publicly—worried that bad actors could use it to attack systems before companies have a chance to fix the vulnerabilities. The move highlights a growing tension: as AI becomes better at spotting weaknesses, companies have to decide whether the risks of releasing such powerful tools outweigh the benefits.