An unsecured database used by a generative AI app revealed prompts and tens of thousands of explicit images—some of which are likely illegal. The company deleted its websites after WIRED reached out.