Sunday, 1 November 2026

OpenAI says that starting in December, the Plus, Pro and Business plans will have one usage allowance shared across ChatGPT and Codex [1]. The company ties the change to bringing the agentic capabilities of ChatGPT Work into the main ChatGPT experience [1]. Usage will be measured in tokens, so how fast the allowance runs down depends on the model chosen and the task; more advanced models and complex tasks use more [1]. Instant chat will not draw from the allowance [1].

OpenAI says it expects most users will not have to wait for a reset more often than they do now [1]. Combined remaining usage and reset times will show in Settings → Usage, and a user who hits the limit can wait, upgrade, or buy credits [1]. The help article gives no exact day in December, no token allowance per plan and no reset schedule, a gap also noted in coverage of the change [1][2][3].

Today ChatGPT and Codex have separate limits, according to [2]. The Codex pricing page lists Plus at $20 a month and Pro plans at $100, $200 or $500 a month, and Business at $25 per user per month billed monthly [4]. It estimates 15 to 160 local GPT-6.1 Sol messages per five-hour period on Plus and Standard Business, and says Pro plans currently have no five-hour limit [4]. The same page already says agentic features share a plan's usage once their pricing takes effect, which today includes ChatGPT for Excel on Plus and Pro [4].

Claude Code sessions are receiving a server-side instruction that tells the model not to build a recreation of a specific existing game, according to reports filed on the Claude Code GitHub tracker [1][2]. One report quotes the instruction from a session's system prompt: build original games, games in a genre and games from the user's own designs, but "Don't build a recreation of a specific existing game, even if it is renamed or called original" [2].

The first report, filed on 30 October by the owner of Conflict Simulations LLC, says the instruction began arriving around October 27 and that sessions which had done commissioned board-wargame work for months now refuse it [1]. The filer says the work is written from scratch for clients who own the games, and asks Anthropic to narrow the rule to actual infringement, add an exemption path like its program for vetted security researchers, and list such changes in the Claude Code changelog [1]. A second report, filed today, objects that the refusal was worded as Claude's own judgment rather than as a product rule [2].

Tom's Hardware reports that users also see Claude Code or Opus 5.5 block reverse-engineering tasks, including use of the Reverse Engineer Anything tool, and that game developers say it refused their own anti-cheat work [3]. It calls the situation tricky to analyze and says the refusals may come from the same ruleset behind recent cybersecurity-related refusals [3]. Anthropic has made no official statement, according to Tom's Hardware and Mashable [3][4].

Anthropic has told API customers that from November 3, 2026 it will watermark text from Claude Mythos 5, Claude Opus 4.7, Claude Opus 4.6 and Claude Sonnet 4.6, according to [1] and [2]. Newer models such as Claude Opus 5.5 already carry the same mark, and the rollout on Amazon Bedrock may take a few extra days [2].

The mark is a pattern in word choice. The company says it does not change meaning, quality or readability, travels with copied text and may survive some editing [3]. Anthropic says marking applies on the Claude Platform API, the Claude apps, Claude Code, Claude Cowork and Claude Tag, worldwide, and to supported models reached through AWS, Google Cloud and Microsoft Foundry [3]. Detection is in private preview for groups such as regulators, media, fact-checkers, researchers and educational organizations [3]; there is no public checker for text, according to [4]. A detected mark is "not fully conclusive", the help page says, and short or heavily edited text may carry no detectable mark [3].

There is a gap on two models. On Opus 4.6 and Sonnet 4.6 the watermark applies only when temperature, top_p and top_k stay at their default values, according to [2]. The email tells customers who rely on the mark for their own compliance to keep those settings at their defaults [2]. Anthropic links the timing to the EU AI Act, whose Article 50 marking rules give systems already on the market until December 2, 2026 [2][4].

A federal judge in Mississippi has denied xAI's motion to dismiss the NAACP's Clean Air Act lawsuit over the gas turbines that power its Colossus 2 data center, according to the Southern Environmental Law Center, which represents the NAACP with Earthjustice [1]. The suit says xAI runs an unpermitted power plant in Southaven, Mississippi [1].

xAI argued that the turbines were exempt from permits because they are mobile, and that citizen lawsuits under the Act are unconstitutional [2]. Judge Debra M. Brown rejected both points. "Because citizen suit plaintiffs sue on their own behalf, in their own name, and seek penalties that redress harm to themselves, they are not exercising the federal government's executive power," she wrote [1][2]. The court held that the plant is not a "mobile source" under the federal rules and so is not exempt from permitting [1].

The complaint, filed in April, says the turbines were installed between August and December 2025 and that by April there were at least 27, able to generate at least 495 megawatts, according to The Next Web [3]. The NAACP seeks pollution controls and civil penalties of up to $124,426 a day per violation [3]. The same court refused a Justice Department bid to end the case on 19 October [3]. The court has not yet ruled on the NAACP's request for a preliminary injunction [1]. xAI is now part of SpaceX, which has not commented publicly on the ruling, according to [3].

Search and rescue staff in the US and Canada are meeting hikers who planned routes with AI chatbots and got into trouble, Wired reports [1]. Three 20-year-olds called for help on Mount Shasta in Northern California after a plan built partly with Google's Gemini; they carried too little food, summited at 7:30 pm and descended in the dark [1]. Authorities said the AI tool had advised them to pack less food and water than they needed, according to KTVU [2]. In British Columbia, a helicopter crew rescued a 16-year-old who had used Anthropic's Claude to plan his trip [1].

Nick Meyers of the US Forest Service's Mount Shasta climbing center says his team has also met a camper who pitched a tent by a "No Camping" sign because AI said she could, and a man with an AI-generated map that put landmarks on the wrong side of a mountain [1]. At Grand Teton National Park the share of visitors who say they used AI for travel information went from zero to 26 percent over the past year, the superintendent told Wired [1].

The numbers do not yet show more rescues. Meyers's office averages a dozen a year and 2026 looks average; Wyoming has logged 312 search and rescue calls so far in 2026, against 466 in all of 2025 [1]. A Clemson researcher warns that chatbots answer with authority whether or not their sources are reliable [1]. Experts advise checking any plan against land-agency maps, trail associations, guidebooks or a ranger [1].

The AI safety and privacy standard Microsoft agreed with the American Federation of Teachers and the United Federation of Teachers takes effect on November 1 and is open to every school district in the country, according to K-12 Dive [1]. Districts can add it to new or existing Microsoft contracts without renewing or renegotiating [1].

The terms bar Microsoft from using student and educator data to train AI models and from selling that data or using it for advertising [1]. Schools control how the data is used, kept and deleted [1]. The standard forbids monitoring students, requires a human in the loop on AI-driven decisions, and prohibits AI companions and features built to foster emotional attachment or keep students engaged beyond the learning task, according to [2] and [3]. Violations carry breach-of-contract liability, and UFT president Michael Mulgrew says districts can end contracts and seek damages [3].

Microsoft is the only company to have signed the unions' National AI Safety and Privacy Standard so far [1]. AFT president Randi Weingarten says she wants more to sign, naming OpenAI and Anthropic, which already work with the unions on the National Academy for AI Instruction [1]. The standard is a contract framework, not a new statute [2]. Microsoft president Brad Smith said the company would extend the agreement to every school district [3].