Each item comes from the AI news of November 14 to 21, 2026 and passed our filter: a firm under 250 staff can use it within 90 days, it touches a cost, a traffic source, or a legal duty, and it comes with one action and a dated primary source. We read every source on November 21, 2026.
OpenAI cuts the price of its top model
On November 17, OpenAI released GPT-6.2 Astra and GPT-6.2 Sol. In the API, Astra’s input and output tokens cost 20% less than GPT-6 Astra’s, and cached input costs 60% less. That puts Astra at $8 per million input tokens, $0.40 cached and $40 output. Sol keeps GPT-6.1 Sol’s prices: $2, $0.10 and $10.
In ChatGPT, both models are open to Plus, Pro, Enterprise, and Edu users. There, OpenAI’s claim is about volume: Sol gets more tasks done on the same paid plan.
This week: if a script or tool of yours calls GPT-6 Astra, rerun one costly job on gpt-6.2-astra and compare the bill. OpenAI’s own tip for paying less: keep the start of each prompt the same and add new instructions at the end, so more of it counts as cached input. Why one model rarely fits every job is in the model mix.
Copilot Business draws the line between included and metered
At Ignite on November 17, Microsoft set out what a Copilot license covers. Chat and Copilot inside Word, Excel and PowerPoint stay in the subscription, and so does Office in Copilot, which brings those apps into the Copilot app. Cowork, Autopilot, and the Work IQ APIs run on Copilot credits, billed by use. So will Copilot Code when it ships in the coming weeks, and Microsoft says more metered services may follow.
For Copilot Business, the plan for firms under 300 users, Microsoft’s billing FAQ says new purchases may come with pay-as-you-go switched on, under a default limit of 4,000 credits per user per month. A user who hits it sees a message and can ask an admin for more. Each spending policy also takes in new metered services automatically, unless an admin turns that setting off.
This week: open Copilot, then Cost management, in the Microsoft 365 admin center. Set a limit that fits your budget, decide whether new services should join the policy on their own, and turn usage billing off if nobody needs Cowork yet. The included features stay either way.
Gemini keeps chats for about two months, even with History off
On November 16, Google split Gemini’s controls in two: one setting for how long your history is kept, another for whether your chats train its models. The change rolls out worldwide in the coming weeks, and there’s a catch: with History off, new chats are now kept for around two months, to answer you, process feedback, and protect users.
Google’s help page adds two details. For adults, training on chat text and attachments is on by default. And on a work or school account, the Workspace administrator sets these controls.
This week: ask who on your team uses Gemini on a personal Google account for work. Have them switch AI training off. Anything about a client belongs on a work account, where your admin sets the rules.







