# Morning digest: Elvis sues Anthropic, Codex reset, and Sweden's AI block

> The owners of Elvis Presley's music sued Anthropic over 274 songs, seeking about $41 million. OpenAI reset Codex limits again overnight.

Published 2026-10-25T06:11:04-07:00 · Daily Digest · Canonical: https://news.littlewolflabs.com/posts/2026-10-25-morning-digest

Morning digest for Sunday, Oct 25. The companies that own Elvis Presley's music have sued Anthropic, the maker of Claude, saying it trained its AI on 274 of his songs without permission and asking for up to $150,000 per song, roughly $41 million. It's a quiet Sunday otherwise, so here's what else is worth knowing: OpenAI reset Codex usage limits again overnight (with some fine print about your refresh day), a developer says he found a bug that drains Claude's five-hour limit faster when you use it through the Agent SDK, Sweden's government offices cut off ChatGPT and Claude from Monday, and a group of more than 1,000 mathematicians who've pledged to keep commercial AI out of their work.

1\. **The owners of Elvis Presley's music sued Anthropic, saying it trained Claude on 274 Presley songs without permission.** The lawsuit was filed Friday, Oct 23, by a unit of Authentic Brands Group, the licensing company that largely controls the rights to Elvis's name, image and much of his song catalog, along with other companies that own parts of his catalog. Billboard reported it first on Saturday. It covers songs including "Can't Help Falling in Love," "Don't Be Cruel," "All Shook Up" and "A Little Less Conversation," and seeks $150,000 per song, the most US copyright law allows for each work when infringement is willful, which comes to about $41 million. "The Elvis Plaintiffs bring this action to enforce these exclusive rights against the defendant that has built its entire business model based on copyright infringement at a scale never seen before," wrote lead attorney Richard Busch, who also represented Marvin Gaye's estate in the "Blurred Lines" case. The suit also says Claude still reproduces songs: according to the filing, when asked recently to recreate specific songs "for a modern audience," Claude did, and admitted the results were "edging past 'inspired by' into reproducing the copyrighted song, even when the words are swapped out." Anthropic didn't respond to Billboard's request for comment on Saturday. AI companies argue that training on existing work is fair use, a legal exception in US copyright law, and courts haven't settled that question yet. *What to watch:* these are claims, not findings. Anthropic already faces suits from music publishers, and Billboard calls this by far the most high-profile case filed on behalf of a single musician, so it adds to the pressure on AI companies to license music rather than fight over it. ([Billboard](https://www.billboard.com/pro/elvis-presley-estate-sues-suno-anthropic/), [Rolling Stone](https://www.rollingstone.com/music/music-news/elvis-presley-rightsholders-anthropic-lawsuit-infringement-1235647682/), [NME](https://www.nme.com/news/music/elvis-presley-rightsholders-sue-anthropic-copyright-infringement-274-songs-3974178))









2\. **OpenAI reset Codex usage limits again overnight, the second reset this week.** Codex is OpenAI's coding agent, and ChatGPT plans include usage allowances for it that refresh every 5 hours, every week, or both, depending on the plan. OpenAI's Tibo Sottiaux posted "Day 20/ Rest and reset" Saturday evening (8:39 PM PT) as part of his 28-day run where each day brings either an improvement or a full reset, added that it would land "By midnight PST," and at 11:06 PM PT wrote: "It is all propagated. Enjoy some mad codexing tomorrow. Fun week ahead too." Day 18, on Thursday night, was also a reset. Not everyone is cheering. Some heavy users point out that a reset can restart your weekly clock, so if you'd saved usage for later in the week, you lose that buffer. OpenAI's help page, updated overnight, confirms how that works: "When a reset restarts a usage period, the new schedule replaces the old one," and "'Global' does not mean every account or every usage limit." *What this means for you:* open Codex's Usage settings (or run `/usage` in the command-line tool) to check your new refresh time instead of assuming your usual day. If you have a saved "banked" reset, it's still in your account; a global reset doesn't use it up. ([@thsottiaux on X](https://x.com/thsottiaux/status/2114200152701575656), [@thsottiaux: propagated](https://x.com/thsottiaux/status/2114237083652882932), [OpenAI Help: How resets work in Codex](https://help.openai.com/en/articles/20001498-how-resets-work-in-codex), [@Im_IrushiK on X](https://x.com/Im_IrushiK/status/2114206023460470899))













3\. **A developer says a bug makes Claude's Agent SDK burn through the five-hour usage limit faster than Claude Code does.** The Agent SDK (software development kit) is Anthropic's toolkit for building your own apps on top of the same agent that powers Claude Code, and when it runs on a Claude subscription, its use counts against that plan's limits. Theo Browne, who makes the T3 Code app, posted Saturday evening (6:25 PM PT): "I confirmed it's a real bug on Anthropic's side. To be clear: this does not affect weekly limits. All Agents SDK usage will drain your 5h limits faster than using Claude Code directly." He said he has reported it to Anthropic and is working with the company on a fix, and he guessed that September's increase to five-hour limits on Pro, Max and Team plans was never applied to the SDK. It started with a Claude Max user who said T3 Code used up a whole five-hour window in about 30 minutes, and a separate GitHub report describes the SDK being blocked while the desktop app on the same account showed 62% of the limit left. Anthropic hasn't commented publicly yet. *What this means for you:* if you run Claude through third-party apps, scripts or `claude -p` and your five-hour window seems to vanish, watch that number specifically. Do the heavy work in Claude Code itself until a fix is confirmed. ([@theo on X](https://x.com/theo/status/2114166439813017873), [@theo: fix in progress](https://x.com/theo/status/2114167316581961932), [@claudeai: September limit increase](https://x.com/claudeai/status/2102435538120691886), [GitHub issue](https://github.com/anthropics/claude-code/issues/104738))













4\. **Sweden's Government Offices will block ChatGPT and Claude on staff computers starting Monday and point staff to Microsoft Copilot instead.** The Government Offices are the civil service that supports Sweden's prime minister and cabinet. In an email to the broadcaster SVT, they said the browser versions of ChatGPT and Claude won't be reachable from work computers from Oct 26, while staff can still use both models through Intric, a separate AI platform the offices use. Copilot stays because it's approved and, according to SVT, its security was raised after a review. The digitalisation director, Mikael Wettercrantz, told staff, in Decrypt's translation: "We have seen several examples of problematic use of open AI services. The examples concern both the sensitivity of the information that has been entered and how the AI-generated responses have been handled." Sweden's Tax Agency and its National Board of Institutional Care already have similar blocks, and the Enforcement Authority is weighing one. *What to watch:* this is about contracts, not which model is smarter. Agencies are approving the AI tool that comes with a signed data agreement, so if you sell or use AI at work, expect the same question about where your data goes. ([SVT](https://www.svt.se/nyheter/inrikes/ai-tjanster-sparras-pa-regeringskansliet), [Decrypt](https://decrypt.co/381763/sweden-government-offices-block-chatgpt-claude))









5\. **More than 1,000 mathematicians have now joined a group that pledges to keep commercial AI out of their work.** The Association for Human Mathematics formed after OpenAI released more than 700 AI-produced math papers at once on Oct 6, a move the group called "not a demonstration of scholarship, but a demonstration of power." Its members page now lists 1,008 members, and 654 of them belong to an "AI-free caucus" that doesn't use AI models in research at all. Every member commits "to not working with commercial AI companies by providing technical labor, mathematical knowledge, consultation, or publicity," and to not publishing AI-generated papers, reviews, lecture notes or books. New Scientist first reported the group crossing 1,000 on Oct 20, and it's drawing fresh attention on X this weekend. "We're not denialists," Steven Sivek of Imperial College London told New Scientist. "We need to figure out how to adapt the profession to this new emerging reality." *What to weigh:* AI labs are using math as their showcase for what their models can do, while a large group of working mathematicians is refusing to help. Treat headlines about AI "solving" math with care until other mathematicians have checked the results and explained them. ([Association for Human Mathematics: members](https://ahmath.org/members), [AHM statement](https://ahmath.org/statements), [New Scientist](https://www.newscientist.com/article/2593621-a-thousand-human-mathematicians-pledge-to-shun-ai-in-their-research/))