The AI Agent Should Recommend. The System Runtime Should Decide.
On August 17, 2026, Google published a reference architecture for building AI agents. The founding assumption of that architecture is that the model cannot be trusted. Shubham Sabo
On August 17, 2026, Google published a reference architecture for building AI agents. The founding assumption of that architecture is that the model cannot be trusted. Shubham Sabo
In June, an AI agent accessed an Australian government health portal it had no authorization to touch. It read files that were not public. It wrote files to an internal server. No
For regulated enterprises governing AI: Is your AI control preventive or detective? Your examiner cares about only one.
Four months. Three labs. Four incidents. The lesson is not that AI is dangerous. You have been putting it in the wrong place in the stack.
If you're running AI Customer Service, or about to, you're in one of two camps: School One is a subscription to a race with no finish line. School two...
Containment is a good metric until your AI issues a refund without verification or a log. Why completion, not containment, is the number that survives an audit.
His thesis: the durable advantage in AI is your own knowledge, not the model you pick. Ashu Garg sharpened it: the asset isn’t the data, it’s the decision trace. In reg
82% of executives feel protected. 88% were already breached. 21% can see what their agents are doing. Three numbers from one report. They describe the same companies, in the same y
The design discipline that will define the next decade of enterprise AI architecture. Picture a renewal flow in a regulated business. A capable model reads the account, understands
Last week, James da Costa and Angela Strange at a16z published Everything, Everywhere is Compliance, arguing that compliance is the biggest and most boring enterprise AI opportuni
