Breaking News: OpenAI Completely Cuts Off Cursor

marsbit2026-08-29 tarihinde yayınlandı2026-08-29 tarihinde güncellendi

Özet

OpenAI has announced it will completely terminate its direct model supply to Cursor, the AI-powered code editor, on November 12. This decision follows the acquisition of Cursor by SpaceX (and thus Elon Musk) in a $60 billion deal two weeks prior. OpenAI cites Musk's history of contractual violations as the core reason, including past instances where xAI (now part of SpaceX) used OpenAI data for model training against terms of service. The move severs Cursor's official bundled access to OpenAI models like GPT. Crucially, it also explicitly excludes access to OpenAI's upcoming, highly capable "Astra" model, which is considered a strategic asset. Developers can continue using OpenAI models within Cursor by supplying their own API key, but this shifts costs from a bundled subscription to a direct, usage-based payment model, effectively raising prices for heavy users. Cursor's CEO confirmed negotiations are ongoing and emphasized Cursor's long-standing relationship with OpenAI, framing the decision as a departure from OpenAI's claimed platform neutrality. The article frames this event as part of a broader industry trend where model providers (like OpenAI and Anthropic) are increasingly cutting off integrated access to their models in tools owned by competitors or entities they distrust. The conclusion is that control over the foundational AI models has become the ultimate source of power, deciding who gets access to the most advanced capabilities.

According to the plan, Cursor's direct access to our model will be completely terminated on November 12th.

As for the next-generation strongest model Astra, it will never be given through official channels.

The reason is simple.

Your boss has a criminal record, I don't feel comfortable.

15 days ago, when Musk bought Cursor for 60 billion, everyone said the same thing—

The Cursor that belonged to global developers is dead.

Sure enough, that was just the opening act.

But Cursor isn't the most tragic. It lost GPT, but still has its own boss's Grok to rely on.

The most tragic ones are the millions of developers who open it every day, and they don't even have a say in this matter.

Seeing OpenAI end its partnership with Cursor, honestly, it's like watching two friends get divorced and then being told I have to choose who to spend the weekend with. I don't want the AI ecosystem to turn into isolated tribes; I want the best models to work together.

Developers only have 75 days left

For developers, there are three things they need to understand today.

The first is the date.

Before November 12th, GPT in Cursor remains unchanged, use it as usual. After that, official direct access is zeroed out.

Counting from today, 75 days.

The second is that the strongest Astra was never intended to be given.

It's not that it won't be given after expiration; it's that it won't be given from now on. This is OpenAI's unreleased next-generation model, also its strongest card.

The third is that two lifelines are still left.

Tibo Sottiaux, nicknamed OpenAI's "godfather," subsequently posted stating:

Developers can continue to use their own OpenAI API Key.

OpenAI's own Cursor IDE extension will also continue to provide access.

So GPT won't be physically erased from the Cursor software; what was cut off is Cursor's direct contract as an official channel.

The difference between the two is:

Before: Cursor bought models in bulk from OpenAI, packaged them into its own subscription, and sold them to you. You paid one monthly fee, which included GPT in your quota. If you exceeded it, you'd top up at Cursor's prices, and if there were issues, you'd contact Cursor.

After: You must register your own OpenAI API Key, enter it into Cursor, and pay OpenAI directly per token. The bill has nothing to do with Cursor, the quota is unrelated to your subscription, and rate limiting and stability are your own responsibility.

It changed from a monthly subscription to self-pay. For heavy users, this is a real price increase.

Just one hour after the announcement, Cursor co-founder and CEO Michael Truell urgently responded.

He said OpenAI models account for about 5% of Cursor user traffic, and the two sides are negotiating, hoping to resolve this matter.

Additionally, this move directly put OpenAI on the spot:

Cursor was one of OpenAI's earliest users. We have collaborated with their team for many years, and we have always trusted their platform to be neutral infrastructure.

Negotiations aren't finished yet, and these words were publicly released.

The calculation is made quite clearly: It's you who are not neutral, not us.

The strongest card will absolutely not be given to Musk

Actually, someone warned about this two months ago.

WIRED's article on July 2nd pointed out two possibilities:

After the acquisition is completed, SpaceX might be unwilling to hand business to Anthropic and OpenAI;

Conversely, Anthropic and OpenAI might also be unwilling to sell models through a product owned by Musk.

Two months later, the latter half came true.

OpenAI was tougher this time than anyone expected.

Because it holds Astra.

On August 7th, OpenAI proactively issued an announcement containing a sentence that made the entire security community's scalp tingle: they cannot rule out the possibility that Astra could reach a "Critical" level of cyber capability.

Translated, this means the model could find and write weapons to breach real-world infrastructure on its own.

For this reason, OpenAI has already locked Astra into an isolated testing environment, restricted network and tool access, added extra encryption to the model weights, and paused a batch of internal projects that didn't meet standards.

Such a card was never meant to be dealt casually.

Let alone when the person waiting to receive it is Musk.

The feud between Altman and Musk is well known.

In April 2023, The New York Times reported that after Musk acquired Twitter, the company violated its contract terms with OpenAI. Twitter is now part of SpaceX.

On April 30, 2026, Musk admitted under oath during testimony that xAI distilled OpenAI's data to train models, violating OpenAI's terms of service. xAI is also part of SpaceX today.

A person who has violated a contract before just bought your largest distribution channel, and you hold a model that can write zero-day vulnerabilities.

Would you give it to them if you were in their shoes?

So for Cursor users, what's lost is not just today's GPT. It's that every time OpenAI becomes stronger in the future, it will have nothing to do with you.

Can Grok fill the gap? Musk himself gave the answer.

On August 15th, the day after the deal closed, Musk expressed his concerns directly at the first all-hands Cursor meeting after the acquisition, in front of over 1,000 new employees.

Grok is not the leader in this market; it urgently needs to catch up.

It's somewhat ironic to spend 60 billion to buy the strongest AI programming tool and admit the next day that you can't replace it.

In 90 days, the AI programming table has been cleared

What's even scarier isn't just Cursor alone.

Arranging the major events in the AI programming circle over the past two years in chronological order reveals the same thing happening repeatedly.

June 2025: Anthropic cuts off Windsurf's Claude access.

August 2025: Anthropic revokes OpenAI's own Claude access.

January 2026: Anthropic blocks xAI. During the same period, the Claude consumer subscription integration in Cursor stops working.

February 2026: SpaceX swallows up the entire xAI, merging it into SpaceXAI.

August 14, 2026: Cursor belongs to Musk.

August 29, 2026: OpenAI cuts off supply to Cursor.

Four years ago, Cursor proved one thing: You don't need to own a model to build a multi-billion dollar company.

Today, the second half of that sentence has been completed—if you don't own a model, you don't have decision-making power.

Acquisitions, blockades, shutdowns appear alternately, with intervals getting shorter.

Cursor's annual revenue is $3-4 billion, with over half of the Fortune 500 using it. A player of this size, cut off just like that.

November 12th.

Remember this date.

References:

https://openai.com/index/our-decision-on-cursor-following-its-acquisition-by-spacex/

Editor: Moses

This article is from the WeChat public account "Xinzhiyuan," author: ASI Revelation

İlgili Sorular

QWhat is the specific date when OpenAI will completely terminate direct access to its models for Cursor?

ANovember 12th.

QWhat is the main reason OpenAI gives for cutting off Cursor from its models, particularly Astra?

AOpenAI cites a lack of trust due to the new owner's (Elon Musk/SpaceX) history of violating contracts and terms of service.

QHow can developers continue to use OpenAI's models in Cursor after the official cutoff date?

ADevelopers can use their own OpenAI API key within Cursor or use OpenAI's own Cursor IDE extension.

QWhat significant change does the OpenAI-Cursor split represent for the AI coding assistant landscape?

AIt demonstrates that companies which do not own their own AI models lack ultimate control and decision-making power, leading to a fragmented ecosystem where major players restrict access to competitors.

QAccording to the article, what did Elon Musk acknowledge about Grok compared to other models shortly after acquiring Cursor?

AHe expressed concern that Grok is not the market leader and urgently needs to catch up.

İlgili Okumalar

Anthropic Eyes 'Training Chips'? Reportedly Considered Acquiring AI Chip Company MatX for $7 Billion

AI giant Anthropic reportedly discussed acquiring AI chip startup MatX for approximately $7 billion to accelerate its in-house chip development, specifically targeting "training chips" for large language models. This move follows OpenAI's recent unveiling of its own inference chip, "Jalapeño," highlighting a growing trend of major AI companies vertically integrating into hardware. However, the MatX deal was ultimately abandoned, with talks shifting toward potential collaboration instead. Anthropic's interest in MatX, a company founded in 2023 by former Google TPU engineers, underscores its strategic push to secure faster and more cost-efficient computing power for its Claude model. The company's broader chip ambitions are further evidenced by its recruitment of key industry veterans, including former Google TPU leader Amir Salek and ex-OpenAI chip engineer Clive Chan, to build an internal chip team. Anthropic is also actively meeting with several other AI chip startups to evaluate different architectures. Despite this significant investment in chip design capabilities, Anthropic reportedly plans to maintain a multi-vendor strategy, continuing its partnerships with major suppliers like Nvidia and Google. Developing advanced chips remains a costly and time-intensive endeavor, making the acquisition of an established startup like MatX an attractive, though currently unrealized, shortcut to gain expertise and potentially reduce long-term costs.

marsbit1 saat önce

Anthropic Eyes 'Training Chips'? Reportedly Considered Acquiring AI Chip Company MatX for $7 Billion

marsbit1 saat önce

Weekly Editor's Picks (0822-0828)

**Weekly Editor's Picks (Aug 22-28)** **Theme:** This weekly digest curates in-depth analysis from the fast-moving information flow, filtering noise to deliver insights. **Macro & Geopolitics** * **US Treasury Strategy:** Wall Street expects potential signals in November regarding future borrowing through more short-term bills and notes, alongside expanded bond buybacks to ease long-term yield pressure. A direct cut in long-term bond issuance is also a rising possibility. * **Gold Outlook:** Goldman Sachs sees fundamental buying, ETF inflows, and options activity driving gold above $4,600/oz. They maintain a year-end target of $4,900, noting potential upside from increased macro hedging demand. Options positions could amplify moves in either direction. **Investment & Crypto** * **Arthur Hayes Interview:** The BitMEX co-founder argues crypto is the primary release valve for central bank liquidity. He predicts ETH could reach $30,000 and sees FLOP potentially surpassing ETH. The Clarity Act is criticized as harmful to US crypto innovation. He views war as the biggest market risk. * **Crypto Leverage Plays:** Analysis of crypto-correlated stocks (MSTR, COIN, etc.) during BTC's 24% weekly surge, ranking their leverage and risk profiles. * **Altcoin Season:** The altcoin market cap surpassed $1 trillion, with 92% of tokens rising. The rally is becoming more fundamentals-driven. * **ZEC & TAO ETFs:** Zcash hit an 8-year high, fueled by Grayscale's progress in converting its Zcash Trust to a spot ETF. A similar "trust-to-ETF" path is noted for Bittensor (TAO). * **Tokenomics Shifts:** Hyperliquid (HYPE) activated a new revenue stream for buybacks. Ethena (ENA) announced a buyback of locked VC tokens and canceled future monthly unlocks, significantly reducing sell-side pressure. **AI & Semiconductors** * **NVIDIA Earnings:** Approaching $100B in quarterly revenue, with growth potentially continuing at 70% next year. Demand is broadening beyond major cloud providers. Supply remains a constraint. * **SK Hynix:** A technical and fundamental analysis following a significant pullback. The company announced a major share buyback. Key risks include competition from Samsung and high stock volatility. **CeFi & DeFi** * **DeFi Picks:** Highlights protocols with strong revenue metrics (UNI, AAVE, JUP, etc.) as potential opportunities. * **Crypto Credit Lines:** Galaxy Digital launched a credit line product allowing users to borrow against BTC, ETH, and SOL portfolios at 8.99% APR, targeting holders needing liquidity without selling assets. **Ethereum & Scaling** * **BitMine's ETH Holdings:** The public company is nearing a 5% stake in all ETH. Analysis discusses the lack of direct network control but highlights concentration risks, regulatory implications, and the company's leveraged bet on a single asset. Fundstrat's Tom Lee suggests buying won't stop at 5% and sees a $10,000 price target for ETH. **Other Notable Topics** * **Meme Coin Drama:** The cycle of rumors, pumps, and dumps around potential "Trump-themed" tokens. * **SOL Burn Proposal:** A vote could significantly increase the daily SOL burn rate. * **Security Incident:** A high-profile Chinese influencer alleged a multi-million dollar crypto scam. * **Weekly Recap:** Key events included BTC reclaiming $80k, speculation about crypto trading on X, and significant altcoin market cap growth.

marsbit1 saat önce

Weekly Editor's Picks (0822-0828)

marsbit1 saat önce

Strive Executive: Rethinking the Bitcoin Price Flywheel

"Strive Executive: Rethinking Bitcoin's Price Flywheel" Bitcoin's maturation process may not follow a simple trend of ever-shrinking returns, as suggested by its long-term power-law trajectory. Instead, a multi-stage "flywheel" effect could emerge, driven by falling volatility. In its early stages, Bitcoin exhibited extreme returns and high volatility, limiting large-scale investment and its use as collateral. As it matures (Stage 2), both returns and volatility decline, improving its risk-adjusted returns. While this seems to point toward diminishing gains, it crucially enhances Bitcoin's appeal to institutional capital and its quality as collateral for loans. Lower volatility allows existing investors to allocate more capital without increasing portfolio risk. More importantly, it significantly increases the amount of debt the system can safely issue against Bitcoin holdings. With shallower potential drawdowns, lenders can extend more credit against the same collateral value, making leveraged Bitcoin accumulation strategies more feasible and resilient. This sets the stage for Stage 3: a self-reinforcing cycle. Improved fundamentals attract more equity capital. Simultaneously, Bitcoin's enhanced collateral status enables the expansion of dollar-denominated credit (e.g., bank loans, bonds) used to acquire more Bitcoin. Fixed Bitcoin supply meets growing demand from both equity and newly created debt, potentially reigniting price acceleration. Thus, the very process of maturation—declining volatility—creates the conditions for a capital and credit flywheel. This could push Bitcoin's USD price to break above its historical power-law trend, analogous to the final, rapid failure stage in a metal fatigue curve where the stressed "material" is the fiat credit system itself.

marsbit1 saat önce

Strive Executive: Rethinking the Bitcoin Price Flywheel

marsbit1 saat önce

İşlemler

Spot
活动图片