As of October 24, 2026. This week AI assistants were given more room to work on their own, with automated checks and owner approval standing in for step-by-step supervision, and a state sued OpenAI over safety claims.
At a glance
| Item | What it is | Who it's for | Our take |
|---|---|---|---|
| Claude Duet | Waitlisted preview of a proactive assistant app from Anthropic | Owners on Claude Pro or Max in the U.S. | Watch it: promising, but still a waitlist preview |
| Claude auto mode default | An automated check replaces per-step approval in most Claude apps | Anyone on Pro, Max or Team plans | Watch it: switch to manual for anything high-stakes |
| OpenAI on Google Cloud | OpenAI models sold through Google Cloud | Businesses already on Google Cloud | Skip it for now: availability is still weeks away |
| ChatGPT locked chats | PIN or biometric lock on selected chats | Anyone sharing a device or account | Try it: useful, but it is not encryption |
| Texas sues OpenAI | State lawsuit over ChatGPT safety claims | Anyone writing AI rules for staff | Watch it: allegations only, early stage |
Anthropic previews Claude Duet
On October 21, Anthropic opened a waitlisted research preview of Claude Duet, a standalone app for iOS, Android and claude.ai. It starts with Pro and Max subscribers in the U.S. It sends a short morning brief from your connected apps, checks them every few hours, and works on tasks in the background on its own cloud computer.
Anthropic says it shows you what it plans to do and waits for your approval before it sends a message, makes a payment, signs a document or deletes anything. Some actions, like moving money between accounts, are off-limits. It can connect to tools such as Gmail, Slack, QuickBooks, Stripe, Gusto and HubSpot, and it shares usage limits with the rest of Claude. Anthropic's announcement has the details, and there is an X post as well.
For a small business: a morning brief across email and invoicing could save an owner some time each week. Connecting payroll and payment tools is a bigger decision than connecting email, so start with the least sensitive app and read every proposed action before approving it.
Claude apps will run in auto mode by default
On October 22, Anthropic said that starting next week, new sessions in most Claude apps on Pro, Max and Team plans will run in auto mode by default. Instead of asking you to approve each step, an automated check reviews actions, lets safe ones go ahead and blocks risky ones. You can switch back to manual approval at any time. Enterprise admins will get an email before the change reaches their organizations, and Anthropic still recommends reviewing actions yourself for high-stakes changes. See the announcement and the X post.
For a small business: if your team uses Claude, decide now who may leave auto mode on and for which jobs. Drafting and research are low risk. Anything touching customer records, money or deletions is a reason to go back to manual approval.
OpenAI models are coming to Google Cloud
On October 21, OpenAI and Google Cloud announced that OpenAI's frontier models and Codex will be offered in Google Cloud's Model Garden and sold through the Google Cloud Marketplace. Companies with Google Cloud spending commitments will be able to apply them to these purchases. OpenAI says the models will also come to Google Antigravity and the Gemini Agent, with general availability planned in the coming weeks. Microsoft remains OpenAI's primary cloud partner. Read the announcement and the X post.
For a small business: this mostly matters to companies already buying through Google Cloud. If that isn't you, nothing changes this week.
ChatGPT adds locked chats protected by a PIN
OpenAI began rolling out locked chats on October 20. You can hide selected conversations from regular chat history, pinned chats and search, then open them with a 4-digit account PIN or, on supported phones, Face ID, Touch ID or device unlock. It is available on Free, Go, Plus, Pro, Business, Enterprise and Edu plans.
OpenAI notes that locking does not add end-to-end encryption or change how long chats are kept. Workspace admins cannot unlock chats or reset the PIN, and resetting a forgotten PIN permanently deletes all locked chats. The help article covers it.
For a small business: it helps when a laptop or phone is shared at the front desk. It does not make it safe to paste sensitive customer data into a chat, and a business admin can't recover a locked chat.
Texas attorney general sues OpenAI and Sam Altman
On October 23, Texas Attorney General Ken Paxton announced a lawsuit against OpenAI and its CEO, Sam Altman. The state alleges OpenAI marketed ChatGPT as safe and reliable while failing to protect Texans, including children, and failing to disclose its risks. The claims are brought under the Texas Deceptive Trade Practices-Consumer Protection Act. These are allegations, and the case is at an early stage. The state's release and an X post are the sources.
For a small business: nothing here changes what you may do today. It is a reminder to write down how your own staff use AI and not to promise customers more than a tool delivers.
What this means for a New Hampshire shop
The pattern this week is assistants that act with less supervision. That is useful for routine work, and it makes the line between what an assistant does alone and what a person approves more important. Money, deletions, signatures and customer messages should keep a person's approval.
If your team is starting to use tools like these, AI cowork training is where we set those lines, job by job. Our post on a one-page AI policy is a good place to write them down.
If you want to talk through it, book a planning call. It's 30 minutes and free.
Sources
- Anthropic previews Claude Duet, a proactive assistant for professionals and small business owners
- Claude apps will run in auto mode by default starting next week
- OpenAI models are coming to Google Cloud
- ChatGPT adds locked chats protected by a PIN
- Texas attorney general sues OpenAI and Sam Altman over ChatGPT safety claims