How Block is building AI-powered security with OpenAI technology
Yesterday, while the market fixated on bigger OpenAI announcements, Block quietly shipped AI-powered security infrastructure across millions of transactions. The setup is deceptively simple: agents watching for fraud patterns in real-time, built on OpenAI tech, running inside one of the world's largest payment networks.
Here's what Block's CTO told the team when they started the project 18 months ago: "Build it like we're the attacker." That framing changed everything. Instead of bolting AI onto existing security rules, they inverted the architecture. The agents simulate threat behavior first, then build defenses against what they find. The system now catches fraud vectors their rule-based stack missed for years—synthetic identity patterns that only emerge across thousands of accounts, timing sequences that look legitimate in isolation.
At QikAI, we've run the same drill with regulated clients. The first question in our Build phase isn't "where can AI help?" It's "where would an adversary aim if they knew we were blind?" That reframe surfaces the real risk surface—usually in workflow gaps between systems, not inside them.
Block's approach carries a second lesson: they didn't start with the hardest problem. They picked a domain where false positives cost credibility but false negatives cost money. Clear feedback loop. Measurable consequence. That's how you build production agents, not pilots that look good in a deck.
The architecture matters more than the model. Block's agents work because they sit inside the transaction path with access to context—user history, device signals, merchant reputation—not outside it making guesses. Compliance and observability were designed in, not added later.
If your AI security strategy still reads "explore use cases," you're already behind an adversary who doesn't pilot.
What's the one security blind spot your current stack can't see because it wasn't designed to look?
#AIAgents #AgenticWorkflows #EnterpriseAI #AIGovernance #ComplianceFirstAI
📩 The full weekly breakdown lands in AI in Telecom, Banking & Retail → https://amplyfy.app/wire/subscribe/18