What about the vulnerabilities you yourself introduce when using AI? or is the assumption that (1) there will be zero or (2) you won't ever open source so no one will know about them?