Great to see @sama and @DarioAmodei to agree on something! Especially after the recent Anthropic Threat Report, which btw was one of the most insane articles I have read recently. If these were reported incidents, you can only imagine how many other exist that we don't know about... Quick TLDR: - Alibaba extracted 151M+ Claude exchanges to help train Qwen. - Moonshot and DeepSeek routed some users to Claude without telling them. - A weapons group used multiple Claude instances as an engineering team, test-fired a guided rocket, failed, and then returned to Claude to understand why. Let's be honest... we don’t hear about such incidents until a someone publishes a report, or someone inside decides to speak publicly... Quite concerning imho, and to be honest with you, a good solution does not exist as of today.