Tech
Satya Nadella has issued a shocking warning to companies using AI
Of all the debates raging about the potential downsides of AI, there is one worry causing the most hand-wringing among AI enthusiasts in Silicon Valley. Their fear is that the giant AI labs that sell proprietary models are somehow acting like Trojan horses.
The concern is that, as startups and enterprises use AI models from labs like OpenAI and Anthropic, the labs gain ever-increasing access to those companies’ most sensitive business information. The model makers can then use that knowledge for themselves, potentially becoming competitors to their own customers. Those issuing such warnings range from VCs like Jason Calacanis to Palantir CEO Alex Karp.
Now, in a surprising blog post published on Sunday, Microsoft CEO Satya Nadella has joined this crowd. Nadella warns that AI users (the “buyers” as he calls them) are paying twice. They knowingly spend for AI token usage but they also, obliviously, hand over valuable data in the process.
“You essentially pay for intelligence twice, once with money, and again with something even more valuable: the proprietary knowledge you must reveal to make that intelligence useful. The better you want the model to perform, the more of that knowledge you have to feed it!” he writes.
Most dangerously, enterprises are literally teaching the models about the nuances of their businesses, he argues.
“Models learn from ‘exhaust,’ the prompts people write, the tools agents use, and especially the corrections people make when the model is wrong. Every correction is distilled into institutional know-how,” he writes.
This is “the kind of knowledge a competitor could never buy,” and yet enterprises are handing it over.
Nadella argues that if AI companies get to freely scrape the internet to train their models, it’s only fair that enterprises get to study — or “distill” — those models in return. “Distillation” is the practice of using a model’s own outputs to learn how it works and to train a new, often cheaper, model based on those insights. In February, Anthropic accused Chinese open source models of sending millions of prompts to Claude as a way to improve their own models, and urged the U.S. government crack down on export controls.
Nadella’s point is that model makers can’t have it both ways. It’s hypocritical for them to freely train on the world’s data while restricting others from doing the same to their models.
“While the great innovation that comes from model providers having fair use rights to train models on public data is needed, I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation,” Nadella writes.
Nadella is particularly concerned when model makers “reserve the right to learn from customer usage and interaction data.”
Nadella’s solution is the kind of thing the CEO of a giant cloud provider would suggest. He wants companies to “retain ownership” of their data, including prompts, feedback, etc. So he’s urging them to build their own “proprietary learning environments” on the cloud (where their data is likely already stored anyway and, conveniently, could mean Microsoft’s cloud, Azure). He also wants companies to build in what he calls “orchestration layers” — essentially, a way to easily switch between AI models from different providers rather than being locked into one. Tools like AI “gateways” that let companies do exactly this have become increasingly popular.
While Nadella never uses the words “open source” as the method for retaining ownership, this is an obvious subtext. Yet, there’s another subtext.
Large companies, many of which still have some of their own data centers in addition to using the cloud, are already moving to open source models installed on their own premises (“on-prem,” in industry jargon). Idit Levine, founder and CEO of Solo.io — which makes networking and security software that helps enterprises manage AI systems — says she’s seeing exactly this shift play out with her own customers. After experimenting with proprietary model makers, they start asking themselves: “Can I take an open source model and run it on-prem? It will do almost 90% of what the big one’s doing. It will cost way less,” she tells TechCrunch. “They understand that, and they can control it.”
Solo.io’s technology was selected last year to be the tech powering the Linux Foundation’s Agentgateway project. Her company counts enterprises like T-Mobile, ADP, and SAP as customers. She sees companies increasingly installing on-premise open source models and sees it as the next big wave in enterprise AI use.
She’s not alone. Vercel (best known as a platform for building and hosting websites, which has recently added AI model-switching tools) and OpenRouter (a company that helps developers route requests across different AI models) are both seeing a surge in traffic to open source models. In fact, open models accounted for 29% of all traffic routed through Vercel’s gateway last month.
With the CEO of Microsoft, a company that has invested in both OpenAI and Anthropic, now openly urging enterprises to be wary of using proprietary models, we’ll bet this trend continues to grow. “In consuming intelligence, you are creating intelligence. And what you create should belong to you,” Nadella writes.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
Tech
Iran abused mobile networks’ vulnerabilities to locate US military in the Middle East, report says
The Iranian government abused well-known vulnerabilities in the global telecoms infrastructure to locate U.S. military personnel in the build-up to the Iran War, as well as in the early days of the conflict, according to Financial Times.
The Iranian government exploited Signaling System 7, or SS7, a set of protocols for 2G and 3G networks that has long been the backbone of how cellular networks connect to each other to route subscribers’ calls and texts around the world, the newspaper reported, citing research by the Mobile Surveillance Monitor, as well as anonymous government officials with knowledge of the spy campaign.
Intelligence agencies have long abused SS7 to track cellphones abroad, which is what happened in this campaign.
Using this technique, Iran was reportedly able to locate U.S. military forces stationed in military bases as well as hotels in Iraq, Bahrain, and other countries in the Middle East, which allowed the regime to strike them. These attacks resulted in several injuries.
Apart from SS7, Iran also abused advertising technology used to serve tailored ads to cellphone users, another well-known surveillance technique that relies on everyday technology.
Tech
Google Images gets a Pinterest-like redesign focused on discovery
Google Images, the tech giant’s image search engine, is taking on Pinterest with its latest redesign that turns the site into a browsable, dynamic gallery of images from across the web. Google is also adding a way for users to create AI images right in Search, as it celebrates 25 years since the debut of Google Images.
Pinterest has long been known for allowing people to browse and save visual inspiration for everything from fashion to home decor. With this redesign, Google is essentially copying that playbook by turning Google Images into a place for discovery and inspiration, and not just search, which could increase users’ time spent on Google platforms, helping boost its ad revenue.
In addition, Google is likely hoping that when users can’t find the image they’re looking for on Google Images or when they want to visualize something, they’ll stay within its ecosystem to create it rather than turn to third-party services like ChatGPT.
After navigating to the redesigned Google Images, users will see a “For You” gallery of images tailored to their interests and browsing history. Like Pinterest, the gallery is designed for continuous browsing, with Google saying it updates in real time with new images.
As users browse, they can save ideas to their “collections,” which will appear as tabs above the main gallery of photos. For example, users can create collections for things like vacation outfit ideas, travel inspiration, and ways to design a reading nook, which they can come back to later.
The redesign is rolling out over the coming weeks on desktop in the U.S. in English. Users need to be signed into a Google Account to try it out, the tech giant says.
As for generating images directly in Search, Google says the feature is meant for moments when you have a highly specific idea for an image that doesn’t already exist online. Google is bringing image generation directly into AI Overviews on Search and will use its latest Nano Banana model to transform a text prompt into a custom visual.
The feature can also help users reimagine spaces and visualize ideas, such as seeing what a room might look like painted red or what a dorm room with a coastal theme could look like.
Image generation in AI Overviews will start to roll out over the coming weeks in English for all regions that currently support image creation in AI Mode, Google says.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
Tech
Meta’s Adam Mosseri says AI token budgets could soon be capped per engineer
In a recent interview, Instagram head Adam Mosseri said he can see a time in the future, perhaps only a year or two, when putting limits on Meta employees’ AI token spend will become necessary.
“I think that you can imagine, at least in a year or two … that the burn rate of a strong engineer might be the same as their salary, or their cost of employment. And in that world, you’re going to probably need to put in some caps,” the Meta executive said, while speaking on Lenny’s Podcast.
AI token spend, a reference to the cost of processing AI prompts and responses, has been a much-buzzed-about subject in recent days. Meta shut down an internal AI token spend leaderboard after AI costs put the company on track for billions of dollars in 2026.
Meta is not alone in rethinking its approach to AI experimentation. Uber also had an AI reckoning after it blew through its 2026 AI coding budget by April. Soaring token costs saw Microsoft cancel Claude Code licenses, consolidating its engineers around its own Copilot CLI tool instead.
Mosseri’s belief, he explained, is that AI token costs will have to be managed just like any other resource, offering an analogy to things like payroll or operating expenditure (OpEx), which is the day-to-day costs of running a business.
“I think of it like…any other resource,” Mosseri said. “I have to decide how to deploy capacity to my different teams because I have a limited number of GPUs and CPUs and storage and RAM etc. I have to decide how to deploy OpEx for labeling budgets across my teams. I have to decide how to deploy payroll for headcount across my teams.”
Token budgets will be the same, he added, noting that the cap per engineer would have to be proportional to the company’s trust in their ability to use the budget in an “ROI-positive” way.
Meta doesn’t currently have token caps for any employee, Mosseri said, but he believes that their use could be healthy in the future. Further down the road, he expects token costs to come down as the AI model makers enter a pricing war to attract people to use their tools over their competitors.
For now, the company has managed to rein in its token costs a bit by shutting down the “silly things” that it was doing, Mosseri noted — like that token spend leaderboard.
“It’s not that hard to build a token incinerator, and that doesn’t create a lot of value,” he said.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
NYT Connections Sports Edition today: Hints and answers for July 15, 2026
NYT Pips hints, answers for July 15, 2026
