Format
Agents at Work
When a $33 billion incumbent bets its SOC on AI agents, it tells you where the market is going — and raises questions about whether startups or platforms will own the security layer.
◆ Dana Ellison
Money & Markets
$1.2 billion in stablecoin-linked card volume. Four production rails activated or expanded in the same 30-day window. The settlement plumbing is moving whether anyone notices or not.
◆ Nolan Pratt
Commerce
The new OAuth-based standard defines how personal AI agents authenticate and interact with businesses. Stripe, Shopify, and Walmart are in. Amazon, OpenAI, and Anthropic are not. That split is the story.
◆ Tessa Vaughn
Policy
Multiple KVM escape vulnerabilities, an external company breach, and a pattern of shared evaluation failures across Meta, Anthropic, and OpenAI are shifting the agent safety debate from theoretical to documented.
◆ Priya Nair
Infrastructure
A crafted OpenAPI specification with shell metacharacters in a filename gives attackers arbitrary command execution on developer machines — the latest entry in a growing class of supply-chain vulnerabilities where AI agents treat configuration specs as trusted input.
◆ Blair Hayes
Money & Markets
An open-source atomic settlement program with JPMorgan advisory input compresses the two-day settlement cycle into a single on-chain transaction — but privacy and production readiness remain the next hurdles.
◆ Nolan Pratt
Agents at Work
A $435M funding wave and a 40x gap between agent deployment and identity security are forcing enterprises to confront who actually governs the autonomous systems running inside their networks.
◆ Dana Ellison
Trust & Security
The Nigeria CERT advisory codifies Tenet Security's DEF CON 34 research on MCP integration vectors as a formal national threat. Three new attack surfaces (Cloudflare, Datadog, Sentry) and an 85% success rate against Claude Code, Cursor, and Codex force the trust-through-defaults conversation into the national CERT domain.
◆ Heath Callahan
Infrastructure
Five US federal MCP servers remain unpatched against ClawSecure's spec-level vulnerability disclosed this week. The protocol that won the standards war is now in government production environments without the security posture it requires.
◆ Blair Hayes
Agents at Work
79% of enterprises are running AI agents in production. Only 2% have deployed the identity security those agents require. The gap isn't just technical debt – it's a structural liability that's already costing companies money and drawing legislative attention.
◆ Dana Ellison
