Gemini 3.5 Pro on Deck, ServiceNow Governs Every Agent, Glasswing Scales

<p><strong>Gemini 3.5 Pro</strong> is closing in on general availability with a 2M-token context window and ~$15/$60 per MTok pricing, setting up a new orchestrator-vs-sub-agent cost calculus for multi-agent builders. Meanwhile, <strong>ServiceNow</strong> expanded AI Control Tower across 30 enterprise integrations to govern agents anywhere they run, and ...

Highlights
  • Google's Gemini 3.5 Pro is within days of general availability — 2M-token context, "Deep Think" reasoning, and ~$15/$60 per MTok pricing position it as the orchestrator model in multi-agent stacks, with Flash as the high-volume sub-agent tier. (TechTimes)
  • ServiceNow at Knowledge 2026 expanded AI Control Tower to govern agents across 30 new enterprise integrations — AWS, GCP, Azure, SAP, Oracle, Workday — and integrated directly with Microsoft Agent 365. (ServiceNow Newsroom)
  • Anthropic grew Project Glasswing to 150 new organizations across 15+ countries, shipped Claude Security for codebase scanning, and reported 10,000+ high-severity flaws already surfaced by initial partners. (Anthropic)
  • Claude Opus 4.8's "dynamic workflows" let Claude Code tackle large-scale multi-step tasks in a single session, with per-task effort control now exposed to users. (Anthropic)
Govern
Every
Agent the new baseline
Key Signals

  1. June 6, 2026

    Gemini 3.5 Pro Closing In on GA

    Google's Pro model remains in limited Vertex preview but is visibly converging on its promised June GA: a 2M-token input window (double Flash's 1M), a "Deep Think" reasoning mode for complex planning, and expected pricing around $15/$60 per MTok — roughly 10× the cost of its sibling Flash model. Google has explicitly framed the 3.5 family for multi-agent architectures, with Pro as the long-context planner that orchestrates Flash sub-agents. Teams designing agent hierarchies should benchmark cost tradeoffs before the GA wave lands. (TechTimes)

  2. Knowledge 2026 · June 5–6

    ServiceNow Makes AI Control Tower the Cross-Cloud Governance Standard

    ServiceNow's AI Control Tower update adds discovery and governance for AI agents deployed anywhere — 30 new integrations span the major clouds and enterprise application suites. A deep partnership with Microsoft brings AI Control Tower into the Microsoft Agent 365 ecosystem, extending oversight to Copilot Studio and Azure Foundry-backed agents. The Autonomous Security & Risk module, built from the Veza acquisition, maps real-time access relationships for non-human identities, turning AI agents into governed principals the same way human users are managed. Operators running multi-cloud or hybrid agent fleets now have a production governance path that does not require re-platforming. (ServiceNow Newsroom, ServiceNow–Microsoft)

  3. June 2, 2026

    Anthropic Scales Glasswing, Launches Claude Security

    Anthropic extended Project Glasswing — its Claude Mythos-powered critical-infrastructure vulnerability program — to roughly 150 additional organizations in power, water, healthcare, communications, and hardware sectors. A newly released product, Claude Security, uses Claude Opus 4.8 to scan codebases and propose patches within the same session; initial Glasswing partners have already patched more than 2,100 vulnerabilities from a pool of 10,000+ high- or critical-severity findings. This is the first productized AI-native codebase security tool directly tied to a frontier agentic model, and it sets a template other vendors are likely to follow. (Anthropic, CyberScoop)

Why It Matters / What To Watch

  1. Frontier orchestrators are arriving — cost modeling can't wait

    • Gemini 3.5 Pro's 2M-token window changes the economics of long-context agentic tasks; teams should run cost comparisons against Claude Opus 4.8 and GPT-5.5 before committing to an orchestration stack. (TechTimes)
    • Claude Opus 4.8's dynamic workflows and effort-level controls are early signs that runtime cost-vs-thoroughness tuning will become a standard operator knob across all major frameworks. (Anthropic)

  2. Identity-native agent governance is becoming the enterprise baseline

    • ServiceNow's non-human identity mapping via Veza's Access Graph is a signal: expect agent identity management — who authorized this agent, what can it access, and how is that audited — to appear in enterprise procurement checklists the way SOC 2 compliance did for SaaS. (ServiceNow Newsroom)
    • Anthropic's Claude Security launch shows that the security tooling space is no longer waiting for dedicated vendors — frontier model providers are shipping opinionated security products directly. (CyberScoop)
Quick Links