https://www.reuters.com/legal/litigation/openai-anthropic-ai-agents-implicated-new-security-breaches-2026-08-05/