Typical AI "security" risk: a bug, not a breach, and you know it'll be forgotten in 6 months after the next shiny exploit is discovered. https://www.reddit.com/user/cyberamyntas