Good morning. If all this AI stuff ever gets to be too much, just tune into the NBA playoffs (go Knicks!).

Way less stressful. Yeah.

Anyway, we’re diving back into DeepSeek today, and the potential security risks it poses.

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

In today’s newsletter:

• ⚕️ AI for Good: Cystic Fibrosis 

• 💰 Heading into tech earnings, here’s what Wall Street is looking for

• 📊 OpenAI’s internal benchmark numbers don’t match Epoch’s 

• 🚨 DeepSeek unmasked: A ‘weapon’ in China’s arsenal

AI for Good: Cystic Fibrosis 

Years before ChatGPT came out, researchers at the Cambridge Center for AI in Medicine (CCAIM) were building machine learning tools designed to enable precision medicine. 

One of their earlier targets, way back in 2020, was Cystic Fibrosis.

The details: A genetic disorder that primarily impacts the lungs, Cystic Fibrosis impacts around 100,000 people globally. Only about half of the people living with the disease in the U.K. are likely to live to the age of 50. 

• Cystic Fibrosis, according to CCAIM co-director Andres Floto, is difficult to treat, since it “is often unclear how the disease will progress in a given individual over time, and there are multiple, competing complications that need preventative or mitigating interventions.”

• Those challenges, combined with the availability of a detailed Cystic Fibrosis database in the U.K., made machine learning a promising avenue. 

Over a two-year period, teams of researchers put together a suite of ML-based tools designed to boost precise, patient-specific diagnostics. These include systems designed to predict lung failure based on an individual patient’s personal data, as well as systems that contextualize the progress of an individual’s disease against other health risks and complications. 

Why it matters: Starting with a basis of accurate and timely predictions, clinicians can engage in treatments and interventions designed to mitigate those risks. 

✂️ Cut your QA cycles down from hours to minutes

If slow QA processes and flaky tests are a bottleneck for your engineering team, you need QA Wolf.

QA Wolf's AI-native service supports both web and mobile apps, delivering 80% automated test coverage in weeks and helping teams ship 5x faster by reducing QA cycles to minutes.

With QA Wolf, you get:

✅ Unlimited parallel test runs

✅ 15-min QA cycles

✅ 24-hour maintenance and on-demand test creation

✅ Zero-flake guarantee

The result? Drata’s team of 80+ engineers saw 4x more test cases and 86% faster QA cycles.

No flakes, no delays, just better QA — that’s QA Wolf.

Schedule a demo to learn more

Heading into tech earnings, here’s what Wall Street is looking for

Amid a tense environment of tariffs and trade wars, market volatility has become the norm. 

The tech-heavy Nasdaq is down some 18% for the year — so far — and the S&P 500 is down more than 12%. This retreat has been led by the same Big Tech names that pushed the major indices to such high highs in 2023 and 2024: Tesla (down 40% on the year), Nvidia (down 29%), Apple (down 20%), Amazon (Down 24%), Google (Down 22%), Microsoft (down 14%) and Meta (down 19%). 

Two members of that group — Tesla and Google — are reporting earnings this week. Here’s what investors will be looking for. 

• For Tesla — a company notorious for lofty expectations and missed deadlines — investors are looking for clarity and hard timelines. Noted Tesla bull Dan Ives said that the company needs to provide timeline details around both the launch of a cheaper car and the launch of unsupervised FSD, slated for sometime this summer in Texas. 

• Ives also wants “hard facts” on what Tesla’s autonomy and robotics roadmaps look like over the next year, as well as details regarding CEO Elon Musk’s role in the White House. 

“We view this as a fork in the road: if Musk leaves the White House there will be permanent brand damage … but Tesla will have its most important asset and strategic thinker back as full time CEO to drive the vision and the long term story will not be altered,” Ives wrote in a Monday note. “If Musk chooses to stay with the Trump White House it could change the future of Tesla/brand damage will grow.”

This comes amid significant declines in Tesla vehicle sales. 

For Google, all eyes will be on cloud growth numbers, especially given its plans to spend an additional $75 billion building out cloud and AI infrastructure this year. Added to that are tariff-related concerns over the near-term health of the digital ad market, Google’s primary revenue source. 

Of course, it is highly possible that none of these details matter. The White House’s unpredictability around massive tariffs, export restrictions and trade wars have been weighing heavily on investors’ minds, lately.

It’s not at all clear that positive earnings or solid outlooks will enable investors to look beyond the macroeconomic maelstrom that’s brewing; this is one of those moments where the inherent volatility of tech (due to high capital expenditures and overly excited future outlooks) is more a hindrance than a help. 

At the beginning of the year, the tech sector made up 32% of the S&P 500. As of March, that number has dropped to 29%, a number that is still significantly higher than its historical average of around 17%. 

The “catch-down” scenario that we talked about in the first week of January is playing out as we speak. 

“Investors care about one thing,” Ives wrote: trade “deals starting to be inked.”

• A pullback: Wells Fargo analysts said in a Monday note that Amazon has paused some of its leasing agreements for new data centers, a move that reads quite a bit like Microsoft’s own data center pullback. It’s a sign that macroeconomic concerns might be inciting a reduction in spending, and it is a potential indication of dark times ahead for big AI names like Nvidia.

• The ‘cancer’ of cyberscam: A UN report this week found that the criminal networks behind the multi-billion-dollar cyberscam industry have expanded into a vast and sophisticated industry. “It spreads like a cancer,” said Benedikt Hofmann, UNODC acting regional representative for Southeast Asia and the Pacific. “Authorities treat it in one area, but the roots never disappear; they simply migrate.”

• Huawei readies new AI chip for mass shipment as China seeks Nvidia alternatives, sources say (Reuters).

• How Big Tech hides its outsourced African workforce (Rest of World).

• Instagram is using AI to find teens lying about their age and restricting their accounts (TechCrunch).

• DOGE is building a master database to surveil and track immigrants (Wired).

• Trump ramps up attacks on Powell, demands ‘loser’ Fed chair lower rates ‘NOW’ (CNBC).

OpenAI’s internal benchmark numbers don’t match Epoch’s 

When OpenAI unveiled o3 last year, the startup said that the model boosted the state of the art on Epoch’s frontier math benchmark to 25% from just 2%. 

But on tests conducted on the recently released model by Epoch, the new number is 10%, a pretty significant difference. o4-mini-high, meanwhile, scored a 17%. 

What happened: What we’re likely seeing here — I say ‘likely because OpenAI isn’t transparent about internal workings — is that the model OpenAI tested on the benchmark isn’t the same as the model that was publicly released. 

• “The difference between our results and OpenAI’s might be due to OpenAI evaluating with a more powerful internal scaffold,” Epoch wrote. 

• The Arc Prize Foundation confirmed this difference in internal models, saying last week that “the released o3 is a different model from what we tested in December 2024 … all released o3 compute tiers are smaller than the version we tested.” 

The difference in capability, according to OpenAI, is the result of optimization for greater cost efficiency. 

It’s another reason, beyond the fact that benchmarks are next to useless when training data is unknown, and that benchmarks rarely test what they purport to test and often don’t carry over to real-world applicability, not to trust benchmark results. 

Both Meta and xAI have been criticized under similar conditions. 

And Epoch, the benchmark company in question here, was criticized earlier this year for releasing initial o3 benchmark numbers before disclosing that it had received funding from OpenAI. 

The companies aren’t straight-up lying about model capabilites; instead, we’re talking about cooking measurements, something that’s compounded by a lack of clear transparency from the developers in question.  

DeepSeek unmasked: A ‘weapon’ in China’s arsenal 

Since its meteoric rise earlier this year, Chinese AI firm DeepSeek has had governments around the world nervous. 

Countries including Taiwan, Australia and Italy moved to ban DeepSeek from government devices shortly after its launch of R1; a number of U.S. states are pursuing, or have already implemented, similar rules. That’s not counting the legislation that was introduced to Congress to, likewise, prohibit DeepSeek from government devices. 

In furtherance of this position, the House Select Committee on China last week published a bipartisan report that, in their words, exposes DeepSeek as a “national security threat to the United States.” 

The details: Beyond claims of “censorship by design,” the report’s real concern deals with issues of surveillance and data privacy. 

• The report claims that, through DeepSeek’s app, the company “funnels Americans’ data to the PRC through backend infrastructure connected to a U.S. government-designated Chinese military company.”

• DeepSeek states in its privacy policy that all the data it collects is stored in servers in China, where, according to the report, it is “subject to the country’s sweeping cybersecurity and intelligence laws, which compel companies to share data with state authorities.”

The report cites a February AP News article that details obfuscatory code in DeepSeek’s website that sends user information to China Mobile, the state-owned telecommunications company that, in 2019, was barred from doing business in the U.S.

“The DeepSeek website and app acts as a direct channel for foreign intelligence gathering on Americans’ private data,” according to the report. “With its direct ties to China’s security and surveillance infrastructure and its unchecked data collection practices, it can function as an open-source intelligence asset feeding American user data into an adversarial system.”

The report’s other concern has to do with the Nvidia chips that DeepSeek obtained despite increasingly strict U.S. export controls; the Committee said it sent the following letter to Nvidia “demanding answers” regarding its Chinese sales. 

"DeepSeek isn’t just another AI app — it’s a weapon in the Chinese Communist Party’s arsenal, designed to spy on Americans, steal our technology, and subvert U.S. law," Chairman John Moolenaar, R-MI, said in a statement. 

In many ways, this report mirrors the political discourse around the TikTok ban, which is almost identically concerned with U.S. user data gathered by TikTok that might be made accessible to the Chinese government. 

The problem with this, at least when it comes to data privacy and security, is in treating a symptom rather than the disease. 

The one thing the entire world cares about more than anything else is data. There exists an entire industry — data brokerage — whose sole goal is the collection and sale of that data. It is currently valued at $434 billion, and is set to surpass $600 billion by the end of the decade. 

And that’s all legal; I’m not counting the proliferation of hackers who steal and sell data on the dark web. 

In other words, as long as other companies — xAI, X, OpenAI, Google, Microsoft, Snapchat, car companies, data brokers, etc. — continue collecting vast quantities of user data, stored in unsecure cloud environments and/or sold to third parties, the Chinese government will be able to get its hands on U.S. data.

🤔 Your thought process:

Selected Image 1 (Left):

• “Hands in 2. Fingers and fingertips in particular. Ew.”

Yeah kind of horrifying if you look close. Don’t know what happened there.

Selected Image 2 (Right):

• “The tool in Image 1 didn't look right to me; guess I have some learning to do about how violins are made!”

💭 A poll before you go

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

We’ll see you in the next one.

P.S. Enjoyed reading? Take The Deep View with you on the go! We’ve got exclusive, in-depth interviews for you on The Deep View: Conversations podcast every Tuesday morning. Subscribe here!

Here’s your view on alien life:

Fully half of you have absolutely no doubt that, somewhere out there, life exists.

15% think it’s unlikely; 13% don’t think so and 18% aren’t sure.

And now I wonder if you think they’re friendly …

Absolutely:

• “While the conditions for life are indeed rare, the odds wholly favor other life in the universe.”

Absolutely:

• “It is just a question of probabilistic evidence.”

If you want to get in front of an audience of 450,000+ developers, business leaders and tech enthusiasts, get in touch with us here.