Overview
- Regulatory capacity came under fire Tuesday as a global study by the Cambridge Centre for Alternative Finance found only two in ten supervisors use advanced AI and just 24% collect data on industry AI use, with 43% having no plans to start within two years.
- Enterprise control gaps widened in Rubrik ZeroLabs’ survey, where only 23% of IT managers reported full control of their agents and 86% said agent growth will outrun security guardrails within a year, with most lacking the ability to roll back bad agent actions.
- Google used its Cloud Next keynote to position Gemini Enterprise as an agent control plane with an agent registry, a tools and skills catalog, and an agent gateway with identity and policy enforcement, plus an Agentic Data Cloud for business context and cross‑cloud queries.
- SAS introduced a Viya MCP Server that lets any AI agent call its governed analytics via the Model Context Protocol, extending bank‑grade fraud models and supply‑chain optimizers to external agents without bypassing SAS’s audit and policy controls.
- Anthropic’s Claude Mythos remains in a restricted consortium after showing it can chain software flaws into working exploits, with reports of limited unauthorized access underscoring calls for zero‑trust agent identities, faster prioritized remediation, and autonomous defensive agents that act at machine speed.