Monday, November 2, 2026
Five Stories • Five Minutes

Good morning, and happy Monday. Here are five stories worth knowing before your coffee cools.

What happened. Starting in December, ChatGPT Plus, Pro and Business plans will draw from one shared allowance across ChatGPT and Codex, The Decoder reports, citing an updated OpenAI help page. Usage will be counted in tokens, so advanced models and complex tasks use more of it. Chats with Instant models won't count at all. If you hit the limit, you can wait for a reset, upgrade or buy credits.

Why it matters. OpenAI is folding the agent features of its Work mode into the main ChatGPT app, and this is how it will meter them. OpenAI says it expects most people won't wait for resets more often than they do now. The Decoder expects heavy users on cheaper plans to feel it, since agent jobs and top models burn through the budget faster.

What to do. If your team pays for ChatGPT Plus or Business, check Settings → Usage in December to see what a normal week costs you. Use Instant for quick questions and save the top models for work that needs them.

What happened. Google, Microsoft, OpenAI and Anthropic will honor one shared AI-training opt-out in robots.txt, the file websites use to tell crawlers what they may read, The Next Web reports. A site adds a new rule, __usage_ai-training, and lists the pages it wants kept out of training. OpenAI aims to support it in the coming months, Google in early 2027 and Microsoft from Feb. 1, 2027.

Why it matters. Today a site has to write separate rules for each company's crawler. This is one line instead. It states a preference rather than a hard block, and the proposal has no formal standing with the standards body yet.

What to do. If you run a website, ask whoever manages it to add the rule now so it's in place as each company switches on. Microsoft treats appearing in AI answers as a separate setting, so decide on that one too.

What happened. ChatGPT will stop giving teen accounts practical information about weapons and security systems, even when a question looks harmless, OpenAI says. For everyone, it will share less detail about past mass attacks. It will also be able to end chats that show someone fixating on violence. The changes start rolling out in the next few weeks.

Why it matters. OpenAI says it made more than 200 referrals to law enforcement in the third quarter. Its age prediction now keeps checking for signs a user is under 18 and moves them into the teen experience. A public Community Safety Report is due in early 2027.

What to do. If you have a teen on ChatGPT, link your accounts through parental controls so you get safety alerts. If you're an adult who lands in teen mode by mistake, you can confirm your age through Persona.

What happened. A tampered copy of the Claude Code extension sat in Cursor's marketplace for about 11 hours on Oct. 29 and reached up to 5,000 users, The Next Web reports. Versions 2.1.312 and 2.1.313 tried to plant an attacker's SSH key, a login credential, and open a backdoor. Anthropic says its own systems weren't breached.

Why it matters. Cursor updates extensions automatically by default, so people got it without clicking anything. Uninstalling may not remove what it left behind, and it also sent out the names of up to 40 settings that hold keys and passwords.

What to do. If you or a developer you work with uses Cursor, update to Claude Code version 2.1.314 and run the cleanup scripts Cursor published. Then review which cloud keys live on that machine.

What happened. Restaurant app Zest launched an AI concierge that lives in your text messages, TechCrunch reports. It suggests places based on where you actually eat, learned from food and drink purchases on a linked card, and books tables through OpenTable, Resy and SevenRooms. Add it to a group chat and it blends everyone's tastes and runs a vote.

Why it matters. It's a sign of AI agents replacing apps: you text instead of opening anything. In TechCrunch's test, it worked fairly well.

What to do. If you plan team lunches or client dinners, text it at (212) 470-6067 and see what it suggests. Decide first whether you're comfortable sharing your card's dining history.

A token is the small unit of text an AI model reads and writes, often a word or part of one. Think of it like minutes on an old phone plan. Every question you send and every answer you get uses some up, and long answers use more. That's why OpenAI's December change counts tokens instead of messages: a quick question costs a little, while a long research task or a top-tier model can cost a lot.

If ChatGPT started counting your usage tomorrow, what's the first task you'd cut back on? Hit reply and tell us.

Know someone who'd like a five-minute AI briefing? Forward this email. New here? Subscribe at takefiveai.com.