Good morning. The New York Times sent a cease-and-desist letter to Perplexity, asking that the AI search startup stop using its content.

This marks the second AI entanglement for the Times, which sued OpenAI last year for massive copyright infringement.

In other news, today, we’re talking about nuclear energy — and the ways in which Big Tech is almost single-handedly bringing it back.

— Ian Krietzberg, Editor-in-Chief, The Deep View

In today’s newsletter:

MBZUAI Research: Evaluating safeguards in LLMs

The jump from the development of Large Language Models (LLMs) to their deployment has proven to be ethically fraught. 

LLMs are known for their propensity to display a range of biases, reproduce copyrighted content and manipulate people. Such models have the capacity to aid cyber criminals and other bad actors in mass cyber-attacks, fraudulent scams and propaganda campaigns. 

In order for developers to mitigate these elements through the enactment of targeted guardrails, they first need to identify them. While closed models, such as ChatGPT, tend to include guardrails that were constructed based upon datasets of known harmful prompts, open-source models tend to lack those guardrails. 

What happened: Researchers from the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) are seeking to bridge that gap, providing a comprehensive dataset for the evaluation of safeguards specifically for open-source models. 

The researchers defined and collected responses that they deemed to be “safe and responsible” which they used to assess the safeguards of several different LLMs. They went on to test automated approaches to assessing these safety mechanisms. 

“Notably, our findings revealed that a suitably-trained small model can effectively perform the evaluation, yielding results that are comparable to those obtained using GPT-4 as an evaluator.” 

To learn more about MBZUAI’s research visit their website.  

If you want to get in front of an audience of 200,000+ developers, business leaders and tech enthusiasts, get in touch with us here.

Semiconductors fall

In reporting third-quarter earnings on Tuesday, chip equipment maker ASML reported lower-than-anticipated sales for 2025, cautioning investors about the remaining weakness in the semiconductor sector. 

The ripple effects of the report — and the firm’s subsequent stock dive — were felt on Wall Street, where a number of major U.S. semiconductors fell in kind. 

Shares of Nvidia retreated more than 5%, a significant pullback from the record highs the firm notched on Monday. AMD, Intel, Arm, Broadcom and Micron fell between 2% and 5%. 

The semiconductor retreat also comes alongside a report out of Bloomberg News that the U.S. is considering applying a cap on export licenses of AI chips to certain countries. 

Google goes nuclear

This week, Google signed an agreement with nuclear energy startup Kairos Power to purchase around 500 megawatts of power, something Google views as an essential step in meeting the rising energy demand of its AI technologies. 

The details: The agreement — the financial terms of which were not disclosed — aims to bring Kairos’ first Small Modular Reactors (SMRs) online by 2030, followed by additional deployments through 2035. 

It’s not clear yet where the reactors will be built, with Kairos saying only that they will be placed in “relevant” locations. 

The context: This comes, as we’ve talked about regularly, in the wake of steadily increasing energy requirements (and carbon emissions) from the data centers that are training and powering AI tech. 

These new energy demands have pushed a number of tech giants, Google and Microsoft included, farther from their net-zero carbon emissions goals, a sacrifice they have deemed worthwhile considering their excitement regarding advancements in AI. It’s worth noting that investments and dominance in AI have served as quite a stock booster from early 2023 on. 

Though data centers currently consume around 2% of global energy consumption, Goldman Sachs expects data center energy demand to grow by at least 160% by 2030; the carbon emissions of the sector, could, in turn, double by that date. (A single ChatGPT query is about 10x as energy-intensive as a single Google search, for comparison). 

In light of this — and the carbon emissions and grid destabilization that has come in hand with it — Big Tech has increasingly begun turning to nuclear energy. 

Just last month, Microsoft signed a deal to restart one of the nuclear reactors at Three Mile Island, the site of the country’s worst nuclear meltdown; in March, Amazon purchased a nuclear-powered data center from Talen Energy. 

Let’s talk nuclear: The SMRs that Kairos offers are, essentially, just smaller than normal nuclear reactors. This makes them more affordable and more flexible — in construction and placement — than traditional reactors. 

Then, of course, there’s the radioactive waste that’s a necessary byproduct of nuclear fission. According to the Nuclear Energy Institute, this waste, once fully cooled in massive pools, gets stored on-site in massive concrete and steel containers called “dry casks.” 

“The energy density of nuclear fuel means that nuclear plants produce immense amounts of energy with little byproduct,” the NEI wrote. “In fact, the entire amount of waste created in the United States would fill one football field, 10 yards deep. By comparison, a single coal plant generates as much waste by volume in one hour as nuclear power has during its entire history.”

Nuclear is not a bad thing, by any means. 

The problem here is that these additional resources are greenifying the massive, additional energy demands coming from the data center; they are not reducing our current consumption.

Clean energy needs to very quickly expand far beyond catching additional requirements to actually reduce total demand. Regularly increasing that demand — provided it is covered by green energy — is not the way forward on that front; we need greater energy efficiencies and a reduction in demand so that clean energy can actually begin to make an impact.

And we need that to happen fast.

What will be interesting to watch, though, is, out of this energy desperation, the ways in which these massive Big Tech investments in nuclear power might actually kickstart a new nuclear energy revolution.

🤔 Your thought process:

Selected Image 2 (Left):

Selected Image 1 (Right):

💭 A poll before you go

Thanks for reading today’s edition of The Deep View!

We’ll see you in the next one.

Here’s your view on Nvidia’s growth:

27% of you think there’s no ceiling on Nvidia’s growth; 27% think it’ll collapse when the bubble bursts.

The rest aren’t really sure.

I guess we’ll have to wait and see.