Five notices, each with its own date and source. Signal: OpenAI says ChatGPT Wallet, which lets an AI agent pay with a saved card, is at limited availability in the US; an FBI advisory includes checks for ConnectWise apps connected to Microsoft 365; AI site builder Solo says it will delete all sites on November 30. Watch: an EU draft impact template and six Copilot models retiring December 14.
Think of this edition as a wall calendar with five entries. One is a wallet that lets an AI assistant pay with your card. One is an FBI advisory with checks for ConnectWise apps connected to Microsoft 365. One is a website builder that will delete every site on November 30. One is a draft EU form for judging how AI affects people's rights. The last lists AI models that GitHub will remove from Copilot on December 14.
Some AI news comes with a clock attached; the dates below are in order.
This edition has three Signal items and two Watch items. Signal means you can act on it this week. Watch means worth knowing; one quick check at most. Every number belongs to whoever published it. The "What it means for you" and "Our take" lines are ours. For the previous edition, see the November 12 edition.
Which dates matter in this edition?
| Date | What happens | Source |
|---|---|---|
| Nov 30, 2026 | Solo deletes all sites and data | Solo FAQ |
| Dec 4, 2026 | Interested organisations encouraged to tell the EU AI Office they want to test the draft | European Commission, Nov 16 |
| Dec 14, 2026 | GitHub deprecates six Copilot models | GitHub changelog, Nov 13 |
| Jan 15, 2027 | EU feedback on the draft template closes | European Commission, Nov 16 |
| Dec 2, 2027 | The Commission's page says deployers will need to comply with this obligation as of this date | European Commission, Nov 16 |
Signal: What is ChatGPT Wallet, and what do OpenAI's pages say about paying and refunds?
OpenAI says ChatGPT Wallet lets a customer's AI agent pay for a purchase with a saved card, after the customer approves it. An agent is software that acts for a person.
Who can use it, and how does a purchase work?
The announcement, dated November 13, 2026, says "We're starting with limited availability to ChatGPT Pro users": subscribers aged 18 and older in the US with eligible US-issued cards. The Help Center article says Wallet is "rolling out gradually", so an eligible Pro user "might not have access yet". No page we read prints a beta, preview or general-availability word, other than a "GA" tag in OpenAI's release-notes listing; the listing gives the same tag to other entries, and its text describes a gradual rollout (see the FAQ). Status as of November 16, 2026: limited availability.
Before each purchase, the user approves with "a security check such as Face ID or a fingerprint scan". The Help article says approval "does not confirm that the merchant has placed your order".
Do the pages agree on approval?
Both the announcement and the Help article describe a biometric check on each purchase; the announcement expects users to delegate purchases entirely to their agent "over time". OpenAI's Wallet terms, effective November 12, 2026, say a user must approve each purchase or "grant ChatGPT the authority to approve each purchase in advance". We state both and resolve neither.
What do the terms say about refunds?
The announcement gives card rewards and disputes two general sentences; the terms are specific, and in our reading they are what a user agrees to, so we carry their wording. The pages differ in that detail. The terms say "The merchant is the seller and merchant of record": order problems and refunds go to the merchant, card billing errors to the card issuer. OpenAI says it "does not decide or fund merchant refunds or card chargebacks", unless the law requires it or OpenAI separately promises otherwise. No page we read states a price or fee for Wallet. We are not lawyers; ask your counsel.
What does OpenAI ask shops to do?
The announcement says "Some sites may unintentionally block purchases through ChatGPT Wallet or other agentic payment products". It lists four areas for businesses:
- Claim your business in ChatGPT and keep a product feed current, as in our November 11 edition.
- "Configure your CDN or firewall to recognize signed requests using Web Bot Auth". Glosses (ours): a CDN delivers your site, a firewall filters traffic, Web Bot Auth lets a bot sign its requests.
- Talk to your payment and risk providers: OpenAI says "risk systems may still flag legitimate purchases made by agents".
- Build a plugin or add WebMCP tools (optional, in our reading).
More on agents that buy: our Visa agentic commerce benchmark decode and our Replit agent teardown.
What it means for you: if you take online orders, ask whoever runs your site and payments how automated purchase requests are treated. We do not advise using it or avoiding it.
Our take: Signal, because OpenAI's own pages list areas a shop can ask about.
Signal: Does the FBI's November 13 advisory on ConnectWise apps and Microsoft 365 apply to you?
Ask whoever runs your IT whether any ConnectWise apps were ever authorised in your Microsoft 365 account. The FLASH says "affected ConnectWise applications" without listing them. If you never used ConnectWise, this item is likely not about you (our reading).
The FBI published a 13-page advisory, which it calls a FLASH, dated 13 November 2026. It is about ARTEX, an open-source tool that uses AI language models to run the steps of a penetration test (an authorised simulated break-in; our gloss) automatically. The FLASH says threat actors have "weaponized" it. It also describes "a supply chain compromise affecting ConnectWise Cloud Backup customers" (an intrusion through a supplier's product; our gloss).
Whose numbers are these?
The FLASH credits two figures to a Microsoft report listed in its notes: the ConnectWise case "enabled access to more than 2,250 downstream tenants" (a tenant is a Microsoft 365 customer account; our gloss), and, in its summary, that some intrusions went "from initial targeting to privileged access in under three hours". The FLASH does not say the FBI counted them. We read Microsoft's October 21 report for these two figures only; it words the first as an actor using the tool "to attempt to sign-in to more than 2,250 tenants", and ties the three-hour figure to "multiple instances" among its examples. In our reading, 2,250 counts sign-in attempts, not confirmed compromises.
What does the FLASH ask organisations to do about ConnectWise?
Three lines of the FLASH's longer recommendations name ConnectWise:
- Review every Microsoft 365 tenant where affected ConnectWise applications are or were installed or authorised, and "search by application ID rather than display name".
- Review activity "starting from at least October 1, 2026" for ConnectWise-related activity.
- "Keep affected ConnectWise application connections disabled until ConnectWise provides formal restoration instructions", and "do not delete enterprise applications as first response".
The FLASH names ConnectWise Cloud Backup in connection with the compromise and does not say ConnectWise did anything wrong. ConnectWise's own advisory (last updated November 4, 2026) says phased restoration of Cloud Backup began October 28, partners re-authorise each Microsoft 365 tenant through a new application, and previously disabled applications must not be re-enabled unless ConnectWise instructs it. We do not say how this fits the FLASH's line about waiting for restoration instructions; ask your IT provider.
This is a separate matter from the November 12 Microsoft 365 outage: the FLASH does not mention the outage, and the outage post we read for our outage article does not mention this tool. ARTEX also appears in our WorkOS decode, a different event.
What it means for you: in our reading, if you use an outside IT provider and Microsoft 365, ask whether ConnectWise apps were ever authorised in your account, whether someone searched by application ID, and how far back the review goes. We are not IT advisers.
Our take: Signal, because the FLASH is official, dated and gives checks an IT provider can run this week. Status as of November 16, 2026: we looked for no FBI update after the FLASH.
Signal: When does the AI website builder Solo delete all sites, and how do you save yours?
Solo is an AI website builder. Its FAQ says "All Solo websites and data will be deleted on November 30, 2026", which is 14 days from today (our arithmetic).
Who runs Solo, and how old is the notice?
The home page notice signs off as "The Solo Team". Solo's Terms of Service and Privacy Policy say the service is "owned, licensed, and operated by Mozilla Corporation". The FAQ is headed "Initial notice: October 1, 2026", so the notice is 46 days old (our arithmetic). Status as of November 16, 2026: the home page says "Registrations are closed" and "Solo will cease operations on Nov 30, 2026".
How do you save a site?
The FAQ says to open Account Settings and click Download Website Data. You get a ZIP file with an HTML export of the site and a CSV file with links to your image assets, "as well as your contact form and newsletter submissions". One catch: "Image source files are not included in the ZIP". The FAQ tells owners to download each image from the CSV's links before the deadline, and says nothing can be recovered afterward.
What about subscriptions and domains?
For Solo Pro or Grow subscribers, the FAQ promises a prorated refund for the unused billing period, "calculated as of October 1, 2026". Domains bought through Solo Domains (not IONOS) must move to Name.com.
The pages we read (the home page, the FAQ, the Terms and the Privacy Policy) do not state why Solo is closing, how many sites it hosts, or how complete the HTML export is as a working site.
What it means for you: in our reading, if a business page, a contact form, a newsletter list or a domain lives on Solo, the export is yours to do before November 30.
Our take: Signal, because the date is hard, 14 days away, and the action is plain.
Watch: What is the EU's draft template for judging how AI affects people's rights, and what are the dates?
The European Commission has published a draft questionnaire and asks for comments. It is meant to help certain organisations that use high-risk AI systems assess, before they start using one, what it could do to people's fundamental rights. The page says it covers "certain deployers" (organisations that use an AI system; our gloss) of high-risk systems "listed in Annex III of the AI Act".
What are the dates?
- The page, dated 16 November 2026, says the draft template is "open for public feedback until 15 January 2027".
- Deployers interested in testing the draft are encouraged to tell the AI Office "by 4 December 2026". Participation is "voluntary and without prejudice to obligations under the AI Act".
- The page says "Deployers will need to comply with this obligation as of 2 December 2027."
What does the page leave out?
The questionnaire is a draft, not a final template and not a new rule; the page presents Article 27 as an obligation under the AI Act. The page says a draft explanatory notice gives guidance on elements such as which deployers need to conduct the assessment; it is a file download we did not open. The page we read states no penalties, no exemptions for small firms, no examples of covered systems and no date for adopting the final template.
What it means for you: in our reading, if you operate in the EU and use an AI system for something that might fall under Annex III, ask your counsel or your AI vendor whether Article 27 applies to you. We are not lawyers; this is not legal advice.
Our take: Watch. The dates are weeks away and the template is a draft.
Watch: Which AI models is GitHub removing from Copilot on December 14?
Status as of November 16, 2026: GitHub's changelog of November 13 says it will deprecate six models "across all GitHub Copilot experiences" on December 14, with a suggested alternative for each:
| Model | Suggested alternative |
|---|---|
| Claude Fable 5 | Claude Fable 5.2 |
| Claude Haiku 4.5 (stays available to individual subscribers on annual plans) | Claude Haiku 5.5 |
| Gemini 3.8 Flash | Gemini 4 Hydrogen |
| Gemini 4 Argon | Gemini 4 Carbon |
| Grok 4.6 | Grok 4.7 |
| Kimi K3 | "To be announced" |
The post gives no reason and no price.
What it means for you: if your team uses Copilot, check which of these models people use, who controls the default, and which workflows and integrations name them.
Our take: Watch, because for Copilot Business and Enterprise GitHub says the alternatives switch on by themselves "unless an administrator has turned off the global default or explicitly disabled the model".
Do this week
Five quick checks; not legal or IT advice.
- If you take online orders, ask whoever runs your site and payments how AI agents' purchase requests are treated.
- Ask whoever runs your IT whether any ConnectWise apps were ever authorised in your Microsoft 365 account, and whether the FBI's checks were run.
- If a site, form, newsletter list or domain lives on Solo, export it, images included, before November 30.
- If you operate in the EU and use AI for something that might fall under Annex III, ask your counsel whether Article 27 applies. Note the dates December 4 and January 15.
- If your team uses Copilot, check which of the six models are selected, who controls the default, and which workflows name them before December 14.
How we know, and what we could not check
We read every source on November 16, 2026 and tested nothing. OpenAI's pages (the announcement, the Wallet terms, the Help Center article and the release-notes listing) were read in a browser; the FBI document as PDF text; the other pages as page text, including ConnectWise's Cloud Backup advisory and, for two figures only, Microsoft's report. Not read: OpenAI's page about dots, its Wallet Privacy Notice and general terms, other ConnectWise pages, the other ten reports the FLASH lists in its notes, Solo's other support pages, the EU draft template and its explanatory notice, and GitHub's supported-models documentation. For help placing agents in your business, see our AI agents work.
Sources
- OpenAI, Nov 13, 2026: Introducing ChatGPT Wallet, read November 16, 2026 in a browser
- OpenAI, effective Nov 12, 2026: ChatGPT Wallet terms, read November 16, 2026 in a browser
- OpenAI Help Center: Making purchases with ChatGPT Wallet (relative date only), read November 16, 2026 in a browser
- OpenAI, Nov 13, 2026: Release notes listing, entry "Introducing ChatGPT Wallet" (tagged GA), read November 16, 2026 in a browser
- FBI, 13 November 2026: FLASH-20261113-001, Threat Overview of ARTEX Autonomous Penetration Testing Framework (PDF, TLP:CLEAR), read November 16, 2026
- ConnectWise Trust Center: Cloud Backup (formerly SkyKick) service interruption advisory, published October 16, last updated November 4, 2026, read November 16, 2026
- Microsoft Security Blog, Oct 21, 2026: report on ARTEX and the ConnectWise supply chain compromise (read for two figures), read November 16, 2026
- Solo home page, shutdown notice (no printed date), read November 16, 2026
- Solo Support: Solo Shutdown FAQ (initial notice October 1, 2026), read November 16, 2026
- Solo Terms of Service (updated June 8, 2026, as printed), read November 16, 2026
- Solo Privacy Policy (updated June 8, 2026, as printed; a previous version dated Feb 14, 2026 is kept on the page), read November 16, 2026
- European Commission, 16 November 2026: Commission seeks feedback on draft template for the fundamental rights impact assessment under the AI Act, read November 16, 2026
- GitHub changelog, Nov 13, 2026: Upcoming deprecation of selected GitHub Copilot models in December, read November 16, 2026