OpenAI Is Cutting Off Elon Musk's Cursor Days After SpaceX Bought It for 60 Billion Dollars

Elon Musk and OpenAI are feuding again, this time over a coding tool millions of programmers use every day. Anthropic spent the week trying to get its AI out from behind a screen and into a working lab. Nvidia backed away from a financing plan that some of its own staff worried looked too close to an antitrust problem. And in India, a sovereign AI stack went live in New Delhi the same week a 12 year old bank account program crossed 59 crore users.

OpenAI cuts Cursor loose over its new owner

OpenAI announced on August 29 that it is ending its partnership with Cursor, the AI powered coding assistant used by millions of programmers, after SpaceX bought the company in a deal reported at 60 billion dollars. Under OpenAI’s proposal, Cursor would lose direct access to OpenAI’s models on November 12.

In a public statement, OpenAI said it could not be confident that SpaceX, run by Musk, would use its technology within the terms of their agreement, pointing to what it described as a pattern of Musk led companies not honoring contracts with OpenAI. Musk and OpenAI have been in and out of court for years, a fight that traces back to his acrimonious departure from the company he helped start.

Cursor’s chief executive, Michael Truell, played down the disruption, saying OpenAI’s models account for only about 5 percent of the traffic running through Cursor’s product today. Anthropic moved fast to fill the gap, pledging expanded Claude model support for Cursor, a reminder of how quickly alliances shift in the crowded market for AI coding tools.

Claude learns to run a lab

Anthropic revealed on August 27 a new standard called the Model Hardware Standard, or MHS, meant to let Claude and other AI agents operate robots, lab instruments, and factory machinery without a human first feeding it a manual. The system stores structured information about how a piece of equipment works, the kind of knowledge that today often lives only in a technician’s head or a stack of paper documentation.

Anthropic said connecting a new piece of hardware to an AI agent normally takes weeks or months and a specialist team. With MHS, the company says that drops to hours or minutes. The standard is launching as a research preview with a small group of partners in science, robotics, and manufacturing, before Anthropic eventually opens it up to any hardware maker that wants to adopt it.

The early pilot numbers were concrete. Claude agents ran a drug discovery experiment at the biotech company Genentech, compressed a multi week imaging project at the Howard Hughes Medical Institute’s Janelia Research Campus into a single day, and pushed the laser stabilization success rate on quantum computers built by QuEra from 58 percent up to 99.3 percent.

Nvidia backs off a deal its own staff didn’t love

Nvidia has paused parts of a financing program that offered smaller cloud computing companies help paying for Nvidia powered data centers in exchange for a cut of the revenue those centers went on to generate. Under the arrangement, called the AI Compute Partnership, Nvidia would take 50 percent of any revenue a cloud provider earned above an agreed hourly rate.

The company launched the program in July and pulled back within two months, according to reports citing internal concerns. Some Nvidia employees warned that taking a revenue share while also gaining influence over which customers a partner could serve looked close to the kind of conduct that draws antitrust scrutiny. Some prospective partners separately balked at giving Nvidia that much say over their business.

An Nvidia spokesperson said the company’s broader approach to financing customer deals is still evolving and remains in place, just not this specific structure. The pause comes days after Nvidia reported the biggest quarter any chipmaker has ever posted, a sign that even at its most dominant, Nvidia is picking its regulatory fights carefully.

India switches on an AI stack built in its own languages

Vice President C P Radhakrishnan launched Gnani Artha on August 28 at his official residence in New Delhi, calling it India’s first sovereign AI stack for enterprises. It was built by the Bengaluru startup Gnani.ai and combines two pieces: Evon 3.3, a 30 billion parameter language model trained natively across eleven Indian languages, and Plexus, an agentic platform meant to connect that intelligence to real business workflows.

The pitch is specific. Indian companies adopting AI keep running into the same three problems: keeping their data inside the country, the cost of running large models at scale, and getting a model that actually works well in Hindi, Tamil, Bengali, and the rest of India’s major languages rather than treating them as an afterthought. Gnani says Evon 3.3 uses about 40 percent fewer tokens than comparable models on Indian language tasks, which lowers the compute bill for anyone running it.

Radhakrishnan called the launch proof that Indian engineers can build frontier technology, not just deploy someone else’s. It is one entry in a much bigger contest, as countries from France to Japan race to field AI systems they do not have to rent from Silicon Valley.

Twelve years, 59 crore accounts, one bank drawer for everyone

India’s Jan Dhan Yojana, the government’s flagship push to give every household a bank account, marked 12 years on August 28. The numbers have compounded steadily since Prime Minister Narendra Modi launched the scheme in 2014 with a promise to end financial exclusion.

The program now covers 59.09 crore accounts, up from 14.72 crore in 2015, according to Finance Ministry data. Deposits sitting in those accounts have grown 12.8 times over the same period to 3.17 lakh crore rupees, and the average balance per account has more than tripled to 5,356 rupees, a sign that people are actually using the accounts rather than opening them once and forgetting them.

Women hold 32.92 crore of the accounts, close to 56 percent of the total, and 45.95 crore, nearly 78 percent, belong to people in rural or semi urban India, the population the scheme was built to reach first. The government also says 41.29 crore RuPay debit cards have been issued through the program, and that 99.92 percent of India’s inhabited villages now have a banking access point within five kilometers. India’s broader Financial Inclusion Index has climbed from 53.9 in 2018 to 67 this year, a slow but real shift in how much of the country has a real foothold in the formal economy.