Recent breaches at Anthropic and OpenAI show how difficult it is to secure advanced AI models as their capabilities—and risks—grow.Full story: https://t.co/RiqHZvjn19— The Information (@theinformation) April 25, 2026