Yesterday's brief put the compromise below the policy layer: when any bearer token passes, the authorization logic above it is theater. Today the frame moves from where the weakness lives to who pays to close it.
The signal that moved is procurement. Palo Alto Networks is now leading its agentic security pitch with AI agent identity and excessive permissions, naming them first-order risks rather than hardening details (Palo Alto Networks). When a security vendor builds its marketing around agent identity, that is the leading edge of a budget line. The Identity Control Surface stops being an architecture diagram and becomes a question a CISO answers in a renewal cycle.
Three product launches this week show the supply side converging on the same answer. IBM's Sovereign Core names agent sprawl as the thing it governs across hybrid environments, inside a four-pillar operating model with governance baked in rather than appended (SiliconANGLE). Google Cloud's Gemini for Healthcare ships a single policy engine with role-based access and native citation trails (HIT Consultant). Adobe's CX Enterprise Coworker builds coordination on MCP and A2A, an interoperability bet that decides whether its agents compose with anyone else's (Adobe Newsroom). Identity scope, policy engine, and interoperable orchestration are each being sold as a product feature, which means each is also a thing an enterprise buyer will now be asked to specify.
The demand side explains why. Capgemini finds 59% of financial firms piloting or deploying multi-agent AI, with only one in ten running agents autonomously across the enterprise, and eight in ten citing technical debt as the binding constraint (LifeHealth). That gap is where identity governance gets purchased. Firms with agents in production and no coordination layer are the exact buyers a Sovereign Core or a Prisma AIRS is written for. The 2026-W43 index holds Governance and Ethics at 79, the highest organizational dimension; these signals show that score cashing out as spend.
For readiness work, the action shifts from writing agent governance policy to pricing it. If identity and permission scope are becoming procurement requirements, the audit question is no longer whether you have a policy, it is whether your current agent deployments could survive a vendor security questionnaire that asks what each non-human identity is permitted to do. Run that questionnaire against yourself before a vendor runs it for you.
Watch item: whether Google Cloud publishes a formal advisory or patch record for the bearer-token authentication weakness documented in the Wiz report. That filing would move this from a researcher finding to a disclosed remediation, and set the template procurement teams cite.
Index Reference · Applied AI Index 2026-W43, the latest published issue (Sunday, October 25, 2026)
WatchGoogle Cloud's expanded AI agent security controls, documented alongside the Wiz finding of an authentication weakness that accepted any bearer token as valid, are moving through enterprise security procurement conversations. The Techwire Asia piece lacks a confirmed publication date, which limits its citation weight, but the underlying Wiz technical report is the kind of third-party vulnerability documentation that accelerates vendor security audits. Track whether Google Cloud publishes a formal advisory or patch record, which would move this from a researcher finding to a disclosed remediation.