Sidekick Orchestration
// The Sidekick Skim · July 3 - 10, 2026

AI This Week: Check Your Microsoft Renewal and Patch Self-Hosted AI

GPT-5.6 became broadly available in three price tiers, Microsoft raised US list prices across most business plans, and a security vendor described an AI-run ransomware operation that entered through an unpatched framework. The practical work is model selection, renewal math, and patch ownership.

In this edition
Read this if you read nothing else
What changed

GPT-5.6 moved from a gated preview to broad availability, Microsoft 365 list prices rose, and a security report described an AI agent carrying a ransomware operation from entry to extortion.

What it means

The premium model is not automatically the right tier. The more immediate business work is understanding your renewal and knowing who patches any self-hosted AI framework.

This week's decision

Benchmark the lowest-cost suitable model, check the renewal date and local price in your Microsoft agreement, and confirm ownership of patches for internal AI tools.

Safe to ignore

The Alibaba and Anthropic dispute and an acquisition involving AI training simulators were interesting signals, but neither required an operating change that week.

Only where it applies
  1. If your Microsoft 365 renewal is approaching15 minutes

    Check the renewal, not the announcement

    Ask your licence owner for the renewal date, current local price, and new price by plan and seat. Compare the total with any separate AI subscriptions serving the same people.

    Limit: The published figures were US list prices. Canadian pricing, negotiated agreements, and renewal timing differ, and existing customers did not all change on July 1.
  2. If your business runs Langflow or another self-hosted AI framework20 minutes

    Confirm patch ownership

    Ask the named technical owner whether the framework and host are current and whether CVE-2025-3248 applies. Record the answer and next review date.

    Limit: If you use a managed service and do not run the framework, ask the vendor rather than attempting to patch it yourself. Verify the current advisory before acting on this archived edition.
  3. If you are considering GPT-5.630 minutes

    Test the cheapest plausible tier first

    Run the same real task on Luna, Terra, and the current model, then compare acceptable output and review time. Move upward only where the cheaper tier fails.

    Limit: High-stakes, customer-facing, or genuinely difficult work may justify Sol. Do not migrate simply because a new flagship became available.
01

The flagship is no longer the default

Operator lens: Relevant if you pay by usage or are choosing a model for repeatable work.

New releaseCompany release; vendor-selected benchmarkGPT-5.6 became broadly available in three tiersIn simple terms: Sol was the premium tier, Terra the balanced option, and Luna the lower-cost tier. The business decision was which level the work required, not whether the flagship won a benchmark.Applies to: Teams selecting models for automation, coding, analysis, or high-volume drafting

Published API pricing per million tokens was Sol at $5 input and $30 output, Terra at $2.50 and $15, and Luna at $1 and $6. On the benchmark OpenAI chose to publish, Sol scored above Fable 5. The more durable signal was the price ladder: Luna's output price was one fifth of Sol's, so routine work no longer needed to inherit flagship economics by default.

Your decision

Benchmark from the bottom of the price ladder upward on your own work. Move to a more expensive tier only when the lower one produces unacceptable results after review. The limit is high-stakes work where a small quality gain is worth materially more than the usage savings.

Sources: OpenAI, TechCrunch

02

Renewal timing matters more than the headline

Operator lens: Relevant if Microsoft 365 is a meaningful recurring seat cost.

PricingOfficial US list pricingMicrosoft raised US list prices across most business plansIn simple terms: Business Basic moved from $6 to $7 per user per month and Business Standard from $12.50 to $14 in US list pricing. Existing customers generally encountered the change at renewal rather than on announcement day.Applies to: Microsoft 365 licence owners, finance leads, and teams reviewing duplicate AI subscriptions

Microsoft's licensing update also listed Office 365 E3 moving from $23 to $26, Microsoft 365 E3 from $36 to $39, and E5 from $57 to $60. Business Premium was listed at $22 with no change, and standalone Teams and Copilot products were excluded. Microsoft noted that pricing varies by country and currency.

Your decision

Ten Business Standard seats at the listed increase equal about $180 USD more per year. Use your actual renewal date and local agreement, then check whether any separate AI subscription duplicates capability now included in the Microsoft plan. The arithmetic is small enough to plan for and not worth panicking over.

Sources: Microsoft Licensing

03

The attack was new. The entry point was not.

Operator lens: Relevant if your organization operates self-hosted AI software or cannot name who maintains it.

Live riskSingle security vendor; unconfirmedA reported AI-run ransomware operation entered through an unpatched frameworkIn simple terms: Security vendor Sysdig described an operation where an AI agent deployed payloads, collected credentials, corrected errors, and encrypted data after entering through a known Langflow vulnerability.Applies to: Teams running Langflow or other self-hosted agent frameworks

Sysdig attributed the entry to CVE-2025-3248, a missing-authentication flaw allowing unauthenticated code execution. It reported more than 600 payloads, 1,342 encrypted configuration items, and credentials harvested from several cloud and AI providers. Every operational detail came from one company selling detection in this category and was not independently confirmed.

Your decision

Name the owner responsible for patches, confirm the affected framework and host are current, and record the next review date. The AI agent may have made the attack faster, but the preventable failure was ordinary patch ownership. If you do not self-host the software, this is a vendor question rather than an internal patching task.

Sources: Sysdig, SiliconANGLE

Evidence note: Evidence status at publication: Microsoft figures were official US list prices, not Canadian or negotiated pricing. GPT-5.6 performance comparisons were OpenAI's selected figures. The ransomware account and operational numbers came from Sysdig, a vendor selling detection for this threat category, and were not independently confirmed. This is an archived edition: verify current prices, model availability, vulnerabilities, and patch status before acting.

// Never miss an edition

Get The Skim every week.

One email, every week, on where AI is heading and what it means for operators.

// One of these decisions applies to you?

Bring one decision. Leave with a next step.

Use the 30-minute call to work through that specific model, workflow, or risk choice.