AI TCH HUM BI Bibek Bhandari AI Agents Don’t Get Hacked. They Get Authorized. The New Vulnerability Class With No CVE In May 2025, a security team in Zurich did something almost boring. They wrote a GitHub issue. That’s just a message on GitHub, the kind…
AI TCH HUM BI Bibek Bhandari Federated Learning in the Real World Is Messy — Here’s What Actually Breaks Most discussions around federated learning assume a clean world. Real systems are anything but.
