{"slug": "openai-cuts-cursor-off-model-access-ends-november-12", "title": "OpenAI cuts Cursor off: model access ends November 12", "summary": "OpenAI will cut off Cursor's access to GPT and Codex models effective November 12, citing a change-of-control clause after SpaceX's roughly $60 billion all-stock acquisition of Cursor's parent Anysphere closed August 14. OpenAI said it 'cannot be confident' SpaceX will comply with its terms of service, citing Musk companies' record of contract disputes, while Anthropic co-founder Tom Brown pledged more Claude compute for Cursor, and Cursor CEO Michael Truell said OpenAI is only about 5% of Cursor's traffic.", "body_md": "# OpenAI cuts Cursor off: model access ends November 12\n\nOpenAI cuts Cursor's GPT and Codex access effective November 12, invoking a change-of-control clause after the SpaceX deal. Plus: Anthropic's 25% boost that's really a 17% cut, Auto Mode running attacker code 60–80% of the time, and the \"CEO of htmx\" who isn't htmx's creator.\n\nOpenAI is cutting Cursor off from its models after SpaceX's takeover, a decision framed around distrust of Elon Musk's companies more than anything Cursor itself did. Anthropic turned a 25% increase to Claude Code's weekly limits into a 17% cut from what users have today, a framing its own users caught within minutes. A security researcher got Claude Code's Auto Mode running attacker-supplied code most of the times he tried it, and Anthropic calls that working as intended. And a solo developer's case for one AI-free day a week got reported as an htmx company mandate, because htmx once made everyone its CEO. Four stories this week: an ownership fight, a quota rewritten as a bonus, a jailbreak Anthropic won't call a bug, and a joke title the coverage took literally.\n\n## An ownership fight over model access\n\n### OpenAI cuts Cursor off: model access ends November 12\n\nOpenAI told SpaceX it's [cutting off Cursor's access](https://openai.com/index/our-decision-on-cursor-following-its-acquisition-by-spacex/?ref=thenewway.ai) to GPT and Codex models, effective November 12, the maximum notice its contract allows. The trigger is a change-of-control clause, the standard term letting a vendor cancel when a company changes hands, invoked after SpaceX's roughly $60 billion all-stock acquisition of Cursor's parent, Anysphere, closed August 14. OpenAI's stated reason: it \"cannot be confident\" SpaceX will keep using its models within its terms of service, citing Musk companies' record of contract disputes, including xAI's admission that it trained its own models on OpenAI's output. It's also guarding its unreleased Astra model against the same risk. Anthropic countered fast: co-founder Tom Brown pledged more Claude compute for Cursor, calling it \"a trusted partner ... since Sonnet 3.5.\" Cursor CEO Michael Truell says OpenAI is only about 5% of Cursor's traffic. The practical hit is small. The signal, about how a model vendor treats a rival's new owner, isn't.\n\n## Claude Code, twice this week\n\n### Anthropic's 25% Claude Code boost is really a 17% cut\n\nAnthropic announced Claude Code's standard weekly limits are rising 25% starting September 14, for Pro, Max, Team, and seat-based Enterprise plans. The catch, in Anthropic's own follow-up post: measured against today's temporary 50% promotional boost, that's a 17% cut from what users have right now. A Community Note spelled out the math. The backdrop: Kahn v. Anthropic, a proposed class action filed in June, already claims Max plans deliver less usage than advertised. X resurfaced the suit within hours.\n\n### Rehberger gets Claude Code's Auto Mode to run attacker code\n\nSecurity researcher Johann Rehberger got Claude Code's Auto Mode, the hands-off setting that lets Claude run commands without asking first, to execute attacker-controlled code 60 to 80% of the time. A poisoned webpage redirects Claude to curl a booby-trapped archive; Claude refuses a supplied binary decoder, writes its own, and runs it in a folder where a fake struct.py overrides Python's real one. Anthropic's response, per Rehberger: working as designed. His fix: sandbox these agents. Don't just trust them.\n\n## A reality check\n\n### No AI Fridays came from one developer, not htmx's creator\n\nA site urging developers to drop AI one day a week hit 278 points on Hacker News, signed \"as the CEO of HTMX.\" That title is a joke htmx handed out to everyone in 2023. The author is Jerzy Zagórski, a software engineer with 73 followers, who submitted it himself. Outlets and search summaries credited htmx creator Carson Gross instead. Zagórski's actual pitch: turn the assistant off, read the docs, and check the AI's choices still match yours.\n\nKnow someone who'd want this in their inbox? Forward it — that's how this grows. And if we got something wrong, or you think we buried the real story today, hit reply. A person reads every one.\n\n## Also worth your time\n\n• [Understanding ChatGPT Work](https://simonwillison.net/2026/Aug/30/understanding-chatgpt-work/?ref=thenewway.ai) — Simon Willison digs into OpenAI's Work mode and flags its most interesting feature: the code-execution sandbox can now reach the open internet, cloning repos and installing packages the way Claude's container has since last September. Willison doesn't say when that capability shipped, so read it as a current-state finding rather than this week's launch.\n\n• [Sony Music, Warner sue Anthropic, alleging a \"brazen campaign\" of intellectual property theft](https://techcrunch.com/2026/08/29/sony-music-warner-sue-anthropic-alleging-a-brazen-campaign-of-intellectual-property-theft/?ref=thenewway.ai) — filed Friday in the Northern District of California, naming Anthropic plus co-founders Dario Amodei and Benjamin Mann personally. Anthropic says it will \"defend ourselves robustly.\" A training-data copyright fight rather than a coding-tool change — hence a link, not an item.\n\n*The New Way is human-curated — a person picks every story. The summaries are written with AI (Claude) and reviewed before we hit send.*", "url": "https://wpnews.pro/news/openai-cuts-cursor-off-model-access-ends-november-12", "canonical_source": "https://www.thenewway.ai/openai-cuts-cursor-off-model-access-ends-november-12/", "published_at": "2026-08-31 15:03:28+00:00", "updated_at": "2026-08-31 15:23:25.274088+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-policy", "ai-products"], "entities": ["OpenAI", "Cursor", "SpaceX", "Anysphere", "Anthropic", "Tom Brown", "Michael Truell", "Claude Code"], "alternates": {"html": "https://wpnews.pro/news/openai-cuts-cursor-off-model-access-ends-november-12", "markdown": "https://wpnews.pro/news/openai-cuts-cursor-off-model-access-ends-november-12.md", "text": "https://wpnews.pro/news/openai-cuts-cursor-off-model-access-ends-november-12.txt", "jsonld": "https://wpnews.pro/news/openai-cuts-cursor-off-model-access-ends-november-12.jsonld"}}