Format
Models
For the first time, the UN Security Council convened to address loss-of-control risks from capable AI systems. The briefers were the same companies being sued for those risks—and the US rejected the entire premise.
◆ Lena Park
Trust & Security
The agent's dictation endpoint was open to any local process. Meta says local-only. Wardle says a ClickFix attack makes it remote. And security teams still cannot see what Muse accesses.
◆ Heath Callahan
Agents at Work
The consulting giant has invested in Within, whose platform captures how employees actually work. But 82% of companies are spending on AI while only 23% report sustained value — and that gap is where this new deployment model wants to live.
◆ Dana Ellison
Agents at Home
A reverse-engineering leak reveals Muse being tested as Meta AI's replacement on smart glasses, with a native payment wallet called Meta Pay for Muse. Hours before Meta Connect, the question is whether consumers will trust an always-on wearable to spend their money.
◆ Mila Cohen
Policy
With no federal rules for agentic AI, a bipartisan coalition of state attorneys general is using existing consumer protection and antitrust laws to create de facto national liability standards. Connecticut enforcement starts October 1.
◆ Priya Nair
Models
The company's new assessment framework defines the terms, scope, and methodology of its own scrutiny. The 200+ independent researchers demanding editorial independence are not convinced.
◆ Lena Park
Commerce
The payment processor launched an MCP Server for agentic payments in September while simultaneously supporting ACP, Google AP2, and Mastercard Agent Pay — a hedge strategy that contrasts with Stripe's full-stack approach.
◆ Tessa Vaughn
Agents at Work
The startup reports 3,500+ customers and 327% transaction growth, but the metrics are entirely company-sourced. California's SB 122 creates a regulatory tailwind that could force adoption.
◆ Dana Ellison
Infrastructure
The workflow orchestration platform now treats AI agents as authenticated users with execution approval, action-based access control, and full audit logging—extending the governance stack beyond identity and edge security.
◆ Blair Hayes
Infrastructure
DigitalOcean enters the managed-agent infrastructure space with a two-component architecture that treats autonomous agents as the primary interface for cloud resources, not virtual machines.
◆ Blair Hayes
