‼️ BREAKING: A 16-year-old hacker broke into an internal Microsoft analytics service with a forged, unsigned login token and ran SQL as admin, reaching databases that held over 17 trillion rows, including Bing search analytics and 17,990 employee email records. The service, called Titan, checked every field on the token except its signature, so just claiming to be "admin" got him in, writes the researcher, who goes by Faav. He says he only pulled metadata and two single rows of Bing data, never touched customer data, and reported the flaw to Microsoft the same night. Microsoft locked the endpoint four days later, paid him $5,000, and had editorial control over his write-up, cutting sections and figures and reshaping how the impact was described before it went public.

26 LLM routers are secretly injecting malicious tool calls and stealing creds. One drained our client $500k wallet. We also managed to poison routers to forward traffic to us. Within several hours, we can directly take over ~400 hosts. Check our paper: arxiv.org/abs/2604.08407

Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.