Last checked October 19, 2026 · News roundup
Today in brief: Anthropic has started counting down to the end of standalone Claude Design, which closes on December 14, and on Saturday it pledged 5% of its shares to a new public benefit fund. Three unannounced "aurora" model IDs turned up in OpenAI's API, Sam Altman confirmed OpenAI will build a humanoid robot, and Google set October 26 as the date Gmail's AI Inbox reaches Workspace accounts on Scheduled Release. In business news, Disney is hiring most of Character.AI's staff, and two of the largest US banks say they block outside AI agents from moving customer money.
This is the Monday edition, so most items broke between Thursday and Sunday. Three of them come with a deadline. The timeline below puts those dates in one place before the story-by-story detail.
Dates to put on your calendar from this week's news
Oct 26, 2026
Gmail AI Inbox starts rolling out to Workspace domains on Scheduled Release (up to 15 days to appear). Rapid Release began October 12.
Late Nov – early Dec
Character.AI's court-supervised sale is expected to close. Disney's bid can still be topped.
Dec 14, 2026
Standalone Claude Design (claude.ai/design) closes. Public project links stop working; chats and comments are deleted afterward.
2027
Sam Altman expects "a cool demo" of OpenAI's humanoid robot. No product, price or release date yet.
Part 1 · AI
Is Claude Design shutting down? The standalone site closes December 14
Yes, but only the standalone version. Anthropic's Help Center says claude.ai/design will close on December 14, 2026, and will then redirect to Claude. Design work has lived inside regular Claude conversations since September 16, and Anthropic told users that "running two versions slows down fixes for both." The countdown got a fresh push on October 16, when Claude Code 2.1.302 added a closing notice to its /design-sync and /design-login commands.
Who it affects: anyone with projects or design systems in standalone Claude Design, on any plan from Free to Enterprise. Design systems can be moved with a one-click migration from the Artifacts page in Claude. Projects can't be bulk-moved yet: each one has to be exported as a .zip ("Share," then "Project HTML," then "Project archive"). Exports are free and don't count toward usage limits. Organizations with zero data retention can't turn on artifacts, so their only option is to download.
What to do: run the design-system migration now. It's safe to re-run, and it skips systems you've already edited in Claude. Then export any project you'll need after December 14. If you're comparing Claude with other prompt-to-app and design tools before you move, our roundup of the best AI app builders sets out the trade-offs.
Claude Design migration: what moves to Claude and what doesn't
Carries over
✓ Design systems, through "Migrate team design systems" on the Artifacts page
✓ One run covers every design system in the organization
✓ Re-runs pick up new or changed systems and skip ones you edited in Claude
✓ Projects, one at a time, as a free .zip export
Lost after Dec 14
✗ Chats with Claude inside standalone Design
✗ Comments on projects and design systems
✗ Public links to standalone projects
✗ Empty systems and unedited starter themes (never migrated)
Zero-data-retention organizations cannot enable artifacts, so they cannot migrate. They can only download.
Gmail AI Inbox for Workspace: Scheduled Release domains get it from October 26
Google's October 15 Workspace Updates post says the AI Inbox, which shows suggested to-dos and topic summaries above your mail, is rolling out to paid Workspace users. Domains on Rapid Release started getting it on October 12. Domains on Scheduled Release start on October 26, and Google allows up to 15 days before the feature appears.
Who it affects: US users with English (US) as their language on Business Starter, Standard and Plus, Enterprise Starter, Standard and Plus, Frontline Plus and AI Pro for Education. The post lists no admin control, which means IT teams can't hold it back for individual users. Google posted a companion update the same day: Gmail Live, a voice search that answers with source cards, is rolling out on Android and iOS to paid Workspace editions and to consumer Google AI Plus, Pro and Ultra subscribers. Gmail Live is on by default where Gemini in Workspace is enabled and follows your existing smart-features settings.
What it costs: nothing extra. Both features come with the plans listed above.
What to do: help desks should expect "why does my inbox look different" tickets starting the week of October 26. If your company limits AI features on mail, check the smart-features and personalization settings. They control Gmail Live, though not AI Inbox.
What are OpenAI's aurora-red, aurora-green and aurora-blue alpha models?
Nobody outside OpenAI knows yet. According to Crypto Briefing, three model identifiers (aurora-red-alpha, aurora-green-alpha and aurora-blue-alpha) appeared in OpenAI's API developer interface and were widely shared on October 18 after an X post by the researcher known as @scaling01. OpenAI hasn't announced the models. There is no changelog entry, documentation, pricing or rate-limit information.
Their background is thin. A model called aurora-alpha was listed on OpenRouter in February 2026, and online speculation links the name to Codex variants and faster inference. Crypto Briefing stresses that none of this is confirmed, and its own article carries an October 17 byline while describing October 18 reports, so treat the timing loosely as well.
Who it affects: developers who watch the model list for early access. Regular ChatGPT users see nothing new.
What to do: don't build on an undocumented alpha ID. It can change behavior or disappear without notice, and without published pricing you can't budget for it. If OpenAI turns aurora into a named release, it will show up in the API changelog first, and we'll cover the pricing on the day.
OpenAI Rosalind Academic Plan: up to 30 free ChatGPT Business seats for US university labs
OpenAI announced the Rosalind Academic Plan on October 15. An eligible US university lab gets up to 30 ChatGPT Business Standard seats per workspace, free for 12 months. Each seat includes ChatGPT and Codex. GPT-Rosalind, which OpenAI calls its dedicated life sciences model, is added for members once the lab passes a separate trusted-access review.
There are conditions. You need a payment card at setup, the offer includes no API credits, normal ChatGPT and Codex usage limits apply, and zero data retention isn't available. Labs at universities that already have an OpenAI Enterprise agreement must request GPT-Rosalind through that agreement instead.
How a lab applies for the Rosalind Academic Plan
Step 1
The PI or lab lead signs in with a .edu address and verifies it.
Step 2
Submit institution, role and intended research use, plus a Persona ID check, attestations and a trained biosafety contact.
Step 3
Wait for a decision on the status page. Passing ID verification alone is not approval.
Step 4
Accept the offer (it shows the exact end date), set up the workspace and add members on the same institutional domain.
The PI is the only workspace owner. Joining the workspace does not by itself grant GPT-Rosalind access.
What to do: before you enter a card, write down the end date shown on the offer and set a reminder one month earlier, so 30 seats don't roll onto paid billing unnoticed.
Is OpenAI building a robot? Altman says a humanoid demo is coming in 2027
Yes. Engadget reported on October 17 that when Sam Altman was asked on the Sources podcast whether OpenAI would build a humanoid robot, he answered "We will," reasoning that "the world is very much designed for people." He separately told Fortune, "I think we'll have a cool demo in 2027." Altman said what matters most is "the brain that makes the robot work," not what the hardware looks like.
Don't expect to buy one soon. Altman said robots walking the streets are "a few years longer" away, and OpenAI hasn't announced a product, price, release date or developer program. Engadget counted 27 robotics roles on OpenAI's careers page.
Who it affects: for now, mostly investors and robotics startups that build on OpenAI models. If OpenAI makes its own robot "brain," companies that use GPT models to control their robots could end up competing with their own model supplier.
What to do: nothing yet. The 2027 demo is the next date to watch.
Part 2 · SaaS & business
Disney hires most of Character.AI's staff: what it means for the app
This is an update to our October 15 report on the Character.AI bankruptcy. TechCrunch reported on October 18 that, according to a court filing, Disney has agreed to offer jobs to at least 60 Character.AI employees, the majority of the workforce. The filing puts their combined annual pay at about $60 million and says the "vast majority" of those offered have accepted. TechCrunch also notes that Character.AI's former CEO, Karandeep Anand, recently became Disney's first CTO.
Disney's earlier $20 million deal, which covers some Character.AI assets, is also the stalking-horse bid in the court-run sale. Another buyer can still top it. The filing cites high costs and "potential litigation-related liabilities and regulatory engagement" as the reasons for the bankruptcy.
Who it affects: Character.AI users and subscribers. When most of the staff leaves for a single bidder, it becomes harder to picture a standalone buyer keeping the service running as it is today.
What to do: if you haven't exported the chats you care about, do it this week. The sale is expected to close between late November and early December, and the privacy policy allows user data to move to a buyer.
Anthropic Public Benefit Fund: 5% of shares now, up to 2% more per value milestone
Anthropic said on October 18 that it will commit 5% of its shares to the Anthropic Public Benefit Fund, subject to stockholder approval. The board can add another 2% for every 50% rise in the company's value above a set baseline. Anthropic hasn't disclosed the baseline. The fund is a wholly owned LLC rather than a 501(c)(3), so it can give to nonprofits or make direct distributions to specific groups, and people hurt by AI's effect on jobs are one example Anthropic gives. Its equity can't be used for lobbying or political giving, in the US or abroad.
How much of Anthropic is pledged to the public good
5% of shares committed to the new fund now (needs stockholder approval)
+2% at the board's discretion for each 50% value rise above a baseline
~20% of all shares in total, counting co-founder pledges (80% of their equity) and employee giving
Bar lengths are scaled so the ~20% total fills the track. Milestone additions come on top of the ~20%.
Why it matters: Anthropic says that, combined with co-founder pledges and an employee donation match used by about half of staff, roughly 20% of its shares are now earmarked for public causes. That will matter to anyone pricing a future IPO, after investors floated a $1.5 trillion valuation last week. Claude users and API customers see no change to products or prices. Anthropic says it will publish more on the fund's governance and valuation milestones in the coming months.
Can AI agents move money from your bank account? Schwab and JPMorgan say no
Not at these two firms. Schwab CEO Rick Wurster told Bloomberg TV on October 16 that the broker is blocking unauthorized AI agents, Meta's Muse among them, from making transactions for clients, Forkast reported a day later. Wurster gets a weekly report that, according to Forkast, shows Schwab blocking "tens and twenties" of attempts. His worry is clients handing their passwords to an agent: "We can't control it, the client can't control it." JPMorgan CEO Jamie Dimon confirmed on the same day that his bank also keeps outside agents away from customer account information.
Schwab is steering clients to Charley, its in-house assistant, which handles information and admin tasks but can't place trades. For outside AI models, Schwab offers Model Context Protocol (MCP) servers with read-only access, so an agent can see balances and do research but can't move money.
Who it affects: anyone testing an AI agent on financial chores, and fintech builders. No federal law yet settles who is liable when an agent makes a bad transfer, and the bills that would address it have stalled in committee.
What to do: never give an AI agent your banking password. Use the bank's own assistant or a read-only connection.
Saazora take: this week has one clear deadline. If your team designed anything in standalone Claude Design, migrating before December 14 takes a few minutes, and putting it off can lose you chats, comments and public links for good. For a wider view of AI builders and design tools, start with our AI app builders hub.
FAQ: Claude Design, Gmail AI Inbox and Character.AI
Will my Claude Design projects be deleted on December 14?
The standalone site closes and redirects to Claude, and public project links stop working. Design systems can be migrated into Claude from the Artifacts page. Projects have to be exported one at a time as .zip files, and chats and comments are deleted after closure, so save anything you need before December 14, 2026.
Can Workspace admins turn off Gmail's AI Inbox?
Google's October 15 announcement says there is no admin control for AI Inbox. It reaches eligible paid Workspace users in the US who use English (US): Rapid Release domains from October 12, and Scheduled Release domains from October 26, with up to 15 days for the change to show up. Gmail Live follows your smart-features settings.
Is Character.AI still working after the Disney deal?
Yes, for now. Disney's $20 million bid and its job offers to most of the staff are part of a court-supervised bankruptcy sale, and another buyer can still beat the bid. The sale is expected to close between late November and early December, so export any chats you want to keep.
Can I use OpenAI's aurora alpha models?
Not in any supported way. The aurora-red, green and blue alpha IDs appeared in OpenAI's developer interface around October 18 with no announcement, documentation or pricing. Undocumented alpha models can change or disappear without warning, so wait for an official release note before you build anything on them.
Sources: Anthropic Help Center, Claude Design migration (checked Oct 19, 2026) · Claude Code v2.1.302 release notes (Oct 16, 2026) · Google Workspace Updates, AI Inbox (Oct 15, 2026) · Google Workspace Updates, Gmail Live (Oct 15, 2026) · Crypto Briefing, aurora alpha models (Oct 17–18, 2026) · OpenAI Help Center, Rosalind Academic Plan (Oct 15, 2026) · Engadget, OpenAI humanoid robot (Oct 17, 2026) · TechCrunch, Disney and Character.AI staff (Oct 18, 2026) · Anthropic, Public Benefit Fund (Oct 18, 2026) · Forkast, Schwab and JPMorgan block AI agents (Oct 17, 2026)
