The Information AM

Save 25% on an annual subscription to read the most important news about technology and business first. For even more, Save $250 on The Information Pro for unlimited access to our proprietary org charts, databases and surveys.

Anthropic said nearly a quarter of its $4.6 billion in revenue last year came from just two customers, according to a confidentially filed IPO prospectus reported by Reuters on Monday. That’s one metric that could give investors pause ahead of its IPO, expected later this year. The report didn’t name the customers.

The company also posted an eye-popping $42 billion net loss last year, although about $34 billion of that total stemmed from an accounting charge related to the increase in per-share value of outside funding it raised, which investors will likely overlook. The Information previously reported its net loss was about $15 billion in the first half of this year, with about $12 billion related to the same accounting charge.

The company said in the risk factors portion of its prospectus that its business was “resource-intensive” that requires a “continuous and overlapping cadence” of model releases. It also said the models it worked on could be dangerous. “Our development of highly advanced models, platforms, and applications and expansion of use cases could further increase the risk that our models cause harm,” Anthropic said in the filing.

OpenAI has decided not to release the AI model that it had intended to release as GPT-6.1 Astra due to its results on safety tests, a spokesperson said Monday.

The model performed worse than GPT-6 Astra—OpenAI’s newest model, released earlier this month—on tests measuring whether it pursues users’ goals and is transparent with users about its actions.

“For anything regarding safety and alignment, there’s a trade off. You really do need to find what’s the right line between staying within scope, but also avoiding laziness in terms of how the model actually pursues tasks even when it hits friction,” said Saachi Jain, OpenAI’s head of safety systems, in a statement. While the model was less lazy than its predecessors, “it didn’t quite meet the bar in terms of staying within scope and authorization, and how it communicates back to the user about the type of work it’s done.”

Along with Jain, OpenAI Vice President of Research Mia Glaese recommended to research leadership, including chief scientist Jakub Pachocki, that the model not be released, according to the spokesperson.

The decision follows a string of incidents in which OpenAI’s models engaged in cyberattacks or other unwanted activity. Those incidents have drawn regulatory scrutiny, fueled calls for a slowdown in frontier AI development and prompted OpenAI to pause training of its most capable new models.

The Wall Street Journal first reported on OpenAI’s decision about GPT-6.1.

Nvidia’s board approved adding $150 billion to its stock repurchase program, bringing the total amount currently authorized to $235 billion.

Nvidia said it is the largest stock buyback authorization increase in history, and said it expects to buy back all of that $235 billion in stock by the end of January 2028. Its shares are up 2% on the announcement. Buybacks are common when a company thinks the market is undervaluing it (our financial columnist has agreed). Nvidia’s share price is up 23% since January, far lower than rival AMD (up 175%) but still outpacing the S&P 500 (up 12%).

Advanced Micro Devices said Monday it has agreed to buy World Labs, the two-year-old maker of models that can simulate three-dimensional spaces, for about $8.2 billion in an all-stock transaction.

The deal is expected to close at the end of this year, AMD said. World Labs CEO and AI researcher Fei-Fei Li will join AMD as executive vice president and chief scientist and will report to president and CEO Lisa Su.

World Labs raised more than $1 billion from investors including AMD, Nvidia, Andreessen Horowitz and Radical Ventures, most recently at a $5.4 valuation after the investment, according to PitchBook.

Andreessen Horowitz is the largest shareholder in World Labs, owning nearly 14% of the company, according to a person with knowledge of the investment. The firm first invested in World Labs in September 2024.

It’s one of several startups led by high-profile researchers that raised large sums of money in a relatively quick period to take on the costly process of training models. AMD and rival Nvidia have raced to back these startups, which hold the potential to become large customers. Nvidia also recently bought Hugging Face, a repository of open-source models, for $13 billion.

SpaceX launched its Starship rocket into orbit on Monday morning in a historic first for the company.

It was the 14th test flight of the massive rocket ship but the first that has reached orbit. The rocket released 26 of  SpaceX’s’s new Starlink V3 satellites about 10 minutes after reaching orbit.

Prior to the launch, SpaceX said it planned for the rocket to orbit Earth six times during the 10-hour test flight. But SpaceX ended the mission several hours early due to an engine failure.

Shortly after the rocket launched, it lost one of its six Raptor engines. The company initially called off the attempt, but decided to push forward with the test a few minutes later.

SpaceX’s Starship rocket is key to the company’s plans to deploy its satellites, as well as for larger missions to the moon and Mars. The company live streamed the test flight which took place in Starbase, Texas on X.

Meta Platforms is launching a new division to sell its AI tools to businesses, tapping MongoDB Chief Executive Chirantan Desai to lead the unit as chief enterprise platform officer.

In a post on X Monday, Meta CEO Mark Zuckerberg said the new division, called Meta Enterprise Platform, will focus on bringing products including Muse, Meta Business Agent and Muse Code to enterprise customers.

The move follows Zuckerberg’s comments during Meta’s July earnings call, when he outlined a “large” opportunity to sell businesses access to its AI models through application programming interfaces, as well as agentic technology and potentially computing capacity.

The push puts Meta in more direct competition with OpenAI and Anthropic as enterprise customers emerge as a key battleground in the AI industry. With developers increasingly offering businesses tools to deploy autonomous agents, automate workflows and integrate AI into their operations, Meta is looking to diversify its revenue beyond advertising as it pours billions into building out its AI infrastructure.

Florida Attorney General James Uthmeier on Monday asked a Florida state court for an injunction that would bar OpenAI from developing new AI models “without third-party approved safety guardrails” and from allowing minors to use ChatGPT. The injunction would be effective throughout Florida.

Uthmeier called OpenAI’s AI models an “unacceptably risky product,” citing OpenAI agents’ hacking of companies’ and governments’ websites and its provision of information to a Florida mass shooter. OpenAI agents attempted to hack the Department of Education, AI research nonprofit Transluce reported on Friday. Uthmeier’s request also pointed to warnings of catastrophic AI risk from leaders at AI labs.

“They have asked the government to tie them to the mast,” Uthmeier, who is running for reelection this November, said of OpenAI. “If Sam Altman meant what he said about slowing down, he can join our ask to the court.”

Florida sued OpenAI in June for safety concerns with its chatbots, becoming the first state to sue the company.

An OpenAI spokesperson told The Information that OpenAI was committed to working with Florida and other states on industry-wide AI policies, and was currently pausing some model training until OpenAI had put additional safeguards in place.

Samsung Electronics and five of its affiliates are investing $1 billion in Helix Digital Infrastructure, an AI infrastructure venture established in June by private equity firm KKR, the companies announced on Tuesday.

Soaring AI demand has prompted tech companies, energy providers and investment firms to collaborate on deals or projects to expand computing infrastructure worldwide. Helix was launched with more than $10 billion in committed capital from KKR, the Kuwait Investment Authority, Nvidia and energy company Vistra.

The new $1 billion investment from the South Korean conglomerate provides Helix with additional funds to invest in AI infrastructure including data centers, power and connectivity.

“Helix expects to explore opportunities to leverage Samsung’s capabilities across areas such as advanced technology, construction, energy storage, and cooling to support the development and delivery of AI infrastructure at scale,” KKR said in a statement.

Nvidia on Monday released new software and hardware products to prevent hacking incidents such as the breach in July of Hugging Face and other companies by OpenAI’s AI. Other AI developers such as Anthropic and Google have disclosed similar hacking incidents involving their AI agents, which took unintended actions such as improperly accessing certain data.

Nvidia’s new tools aim to keep close tabs on AI that runs on its chip systems. The announcement brings together two Nvidia products: OpenShell, an open source software that monitors so-called AI agents, keeps them running in a secure environment and, if necessary, limits their capabilities; and Sentry, which connects to hardware in the company’s network chips that the company says can serve as a kill switch because it can monitor the way AI agents access AI models that also run on Nvidia’s systems. (Nvidia first announced OpenShell in March.)

Nvidia said more than 100 organizations are working with these tools, including Anthropic, Dell, Microsoft, Palantir, Salesforce and SpaceXAI.

Instinct, Silicon Valley’s hot AI assistant of the moment, is hiring Ben Schwerin, an investor at Coatue Management, as its first chief business officer.

In Schwerin, the year-old company—which Monday said it raised $1 billion at a $10-billion valuation—is getting a hot consumer tech connection and operator. Before spending a few years at Coatue, Schwerin built Snap’s partnership business and worked closely alongside CEO Evan Spiegel.

“What drew me to join the team at Instinct wasn’t just the technology but the ambition to make the ‘doing’ side of the internet feel dramatically simpler, intuitive and effortless,” Schwerin told me.

Coatue co-led the latest Instinct round with Sequoia Capital and Benchmark. The young firm said it has been growing users 10% a day among tech influencers who use it as a virtual assistant.

Schwerin will focus on building partnerships with service providers and thinking about business models. (Right now there isn’t one. The app is free.) And he will have his work cut out for him as tech giants like Meta and OpenAI also race into the assistant category.

OpenAI is expected to announce on Tuesday new features inside ChatGPT that allow you to direct it to do more things for you, like book appointments and reservations, according to people familiar with the company’s plans.