Time is running out to prosecute Dr. Anthony Fauci. After May 11, Fauci — the man touted as “America’s doctor” during the pandemic and who led the US response — will be clear of the five-year statute of limitations. Past that date, he can’t be indicted over allegations that he lied to Congress about the origins of Covid or the US government’s role in funding research that may have led to the pandemic.

Six years after a contagion that killed millions and wrecked the global economy, one of the great unanswered questions of Covid is, how did it begin? Today, an overwhelming majority of Americans reject what in 2020 was more or less the official (but outlandish) story: namely, that the novel coronavirus sprang from a wet market in Wuhan, China, and somehow jumped from bats — or pangolins? — to humans. 

This zoonotic theory of Covid’s origins still enjoys the backing of some scientists, but it is bedeviled by questions proponents can’t seem to shake off. For one thing, what was the intermediate animal that supposedly facilitated transmission? Who was Patient Zero? And how was it that the epicenter of the pandemic just happened to be located in Wuhan — one of a few virology labs in China authorized for research on dangerous coronaviruses?

That latter question points to a different hypothesis for Covid’s origins: that it resulted from a lab leak at the Wuhan Institute of Virology, where researchers had been conducting cutting edge “gain-of-function” research, combining different viruses to create dangerous hybrids for the purpose of better understanding future infectious diseases. The lab-leak theory has been endorsed by at least one agency in the US Intelligence Community — the CIA.

Yet Americans and the wider world may never know the full truth unless these questions are subjected to a deep and rigorous examination — and a Fauci trial, conducted to the standard of “beyond a reasonable doubt,” may be the only means left for this. 

The justice system already appears to be tending in this direction. The Department of Justice this week charged David Morens — a former senior adviser to Fauci, then the director of the National Institute of Allergy and Infectious Diseases, part of the National Institutes of Health. The government accuses Morens of conspiring against the United States and destroying records related to Covid’s onset, among other crimes (Morens is presumed innocent until proved guilty).

The Morens indictment mentions “co-conspirators.” Americans must know: was his then-boss, Fauci, one of them? What we already know makes it implausible that he wasn’t. That makes a formal indictment all the more urgent for truth and accountability.

Since the early days of Covid, much of the scientific establishment has been working to discredit the lab-leak theory. As head of NIAID during the pandemic and the face of the government’s response, Fauci was a key player in this effort. Early on, he dismissed as a “conspiracy theory” the notion that Covid might have been created in a lab and leaked by accident. Yet there are reasons to suspect that NIAID under the leadership of Fauci funded the research that may have created Covid.

The case against Fauci, if the Justice Department brings one in the coming days, will likely revolve around charges that he perjured himself in testimony before Congress about this question of research funding and Covid’s origins. At a Senate committee hearing in May 2021, Fauci categorically stated in a tense exchange with Sen. Rand Paul (R-Ky.) that the NIH and NIAID did not fund gain-of-function research in Wuhan. 

“You are entirely and completely incorrect,” Fauci snapped at Paul. “The NIH has not ever and does not now fund gain-of-function research in the Wuhan Institute of Virology.” At a follow-up hearing in July 2021, Fauci got into it again with the Kentucky senator, saying he didn’t retract his May statement about gain-of-function funding. At one point, he jabbed a finger at Paul and shouted, “If anybody’s lying here, senator, it is you!”

But this week’s indictment of Morens for concealing federal records related to the pandemic has revived the debate about Fauci and created new pressure on the Trump DOJ to revisit Fauci’s truthfulness before Congress. 

The records in question have to do with precisely what Fauci has repeatedly denied: gain-of-function research at the Wuhan Institute of Virology — funded by an NIH grant. According to the indictment, Morens pledged to help “Co-Conspirator 1,” described in the indictment as the head of a US-based nonprofit, restore a canceled NIH grant for “Company #1.” The canceled grant was titled, Understanding the Risk of Bat Coronavirus Emergence, and it was canceled amid allegations that Covid came from the Wuhan institute.

The descriptions of Co-Conspirator 1 and Company #1 point directly to Peter Daszak, former head of the now-defunct EcoHealth Alliance, which subcontracted a 2014 NIH grant to the Wuhan lab for gain-of-function research on bat coronaviruses. (Neither Daszak nor EcoHealth Alliance are named in the indictment, but reporting in The New York Times and elsewhere imply that they are indeed the person and entity referred to in the indictment). 

What’s more, the indictment alleges that these co-conspirators gave Morens “illegal gratuities” and wine for his “behind-the-scenes shenanigans.” As Acting Attorney General Todd Blanche put it in a statement: “Morens … allegedly identified an official act that he could perform to ‘deserve’ the gift, which was a scientific commentary in a prominent medical journal advocating that Covid-19 had natural origins.”

To put the dots together: The government’s case against Morens suggests that the NIH did approve a grant that funded precisely the kind of research that could have created Covid, in the very city where Covid first appeared. Once the pandemic broke out and EcoHealth Alliance lost its NIH funding, Morens allegedly tried to help restore this funding in his official capacity as a senior NIH official, but concealed his efforts to do so. It follows that Fauci’s statements before Congress categorically denying such funding were false.

The indictment also corroborates a statement by NIH principal deputy director Lawrence Tabak, who at a 2024 House committee hearing admitted that the US government funded gain-of-function research in Wuhan. Asked directly by then-Rep. Debbie Lesko (R-Ariz.) whether NIH funded gain-of-function research at the Wuhan lab through EcoHealth Alliance, Tabak said: “It depends on your definition of gain-of-function research. If you’re speaking about the generic term, yes, we did.”

Splitting hairs over what exactly constitutes gain-of-function research, or what the accepted definition of the term is, was a recurring feature of Fauci’s sparring in Senate hearings with Paul and other lawmakers during the pandemic. At a Senate hearing in November 2021, Paul brought up the gain-of-function research on bat coronaviruses that EcoHealth Alliance subcontracted to the Wuhan lab through its 2014 NIH grant. Fauci responded by calling Paul’s line of questioning an “egregious misrepresentation,” and insisting that his characterization of the research didn’t meet the NIH’s definition of gain of function, which was formalized by the White House Office of Science and Technology Policy in January of 2017.

Yet Fauci’s own past statements uphold Paul’s definition of gain of function. In a June 2021 letter, Reps. James Comer and Jim Jordan — then the Republican ranking members of the House Committee on Oversight and Reform and the House Committee on the Judiciary, respectively — asked Fauci to reconcile his past statements with his May 2021 assertion under oath that the NIH never funded gain-of-function research at the Wuhan lab. 

The letter quoted Fauci at a 2012 conference on gain-of-function research at which he said: “what historically investigators have done is to actually create gain of function by making mutations, passage adoption, or other genetic techniques, such as reverse genetics.” The letter then quoted the 2014 NIH grant to EcoHealth Alliance, which allowed it to “test predictions of [coronavirus]…transmission…using reverse genetics…” In other words, NIH was contracting research that fit Fauci’s own definition of gain of function. 

“Using your own definition,” wrote Comer and Jordan, “it appears the NIH funded gain-of-function research at the [Wuhan Institute of Virology].” They closed their letter by asking Fauci to “explain the apparent discrepancy in your recent testimony.”

Fauci, who once claimed to personally embody science, has been widely criticized for playing fast and loose with the facts. In addition to his shifting statements about masking and social distancing protocols during the pandemic, he was secretly involved in a widely cited March 2020 paper published in the prestigious journal Nature Medicine on the “proximal origins” of Covid. The paper was designed specifically to discredit the idea of a lab leak. “Our analyses clearly show that SARS-CoV-2 is not a laboratory construct or a purposefully manipulated virus,” the authors stated. “We do not believe that any type of laboratory-based scenario is plausible.”

Fauci mentioned this paper to White House reporters in April 2020 as evidence against the lab-leak theory of Covid’s origin — without revealing that he had been involved in its creation, as emails released in a congressional inquiry eventually revealed. RealClearInvestigations journalist Paul Thacker uncovered that Fauci admitted during a multiday deposition by the House Oversight Committee in 2024 that he had indeed read multiple drafts of the paper, and even emailed with the authors about it. Yet Fauci had previously told The New York Times that he wasn’t sure if he ever got around to reading the paper.

It’s easy to dismiss all this as water under the bridge. The pandemic is over, Fauci is out of the government. In the words of that infamous 2022 essay in The Atlantic, can’t we just declare a “pandemic amnesty” and move on?

In a word, no. As RealClear’s Thacker has noted elsewhere, while “dissembling to the media is not a crime, lying to Congress is illegal.” More than any other single person, Fauci was in a position to shed light on the origins of Covid at the moment the pandemic arrived on American shores. It might have required coming clean about what the NIH and NIAID had been funding in China on his watch, and it might have even required him to resign in disgrace over the scandal such an admission would cause.

But it would have been the right thing to do. In addition to helping government officials and medical providers better understand what we were dealing with, possibly even saving American lives, it would have opened up space for a reckoning about what US tax dollars had been paying for overseas. Later, there might have been a chance at accountability for those responsible for funding reckless gain-of-function research.

 Instead, Fauci dissembled about what he knew, the government’s role, and the part he himself played in all of it. There is still a chance to hold him at least partially accountable. But time is running out.

---

John Daniel Davidson is a senior editor of The Federalist.

johnddavidson

LLM (google/gemini-3.1-flash-lite-preview-20260303) summary:

• Corporate Survival: the entity formerly synonymous with obsolete mobile handsets clings to existence by abandoning consumer hardware for obscurity.
• Market Repositioning: the company survives solely by licensing internal software components to automotive manufacturers who prefer remaining unnamed.
• Operational Scope: proprietary systems are currently embedded within the infrastructure of approximately two hundred seventy five million vehicles globally.
• Hidden Utility: the software operates as a background utility for driver assistance features while remaining completely invisible to the end user.
• Diversification Efforts: reach has expanded into medical robotics and industrial sectors where automation replaces human oversight.
• Financial Reality: quarterly profits have surfaced only after the company essentially pivoted away from its failed legacy consumer brand.
• Stock Performance: market valuation remains a minor fraction of historical peaks despite recent speculative rallies following erratic corporate maneuvering.
• Strategic pivot: internal focus shifted toward backend systems after the realization that consumer-facing infotainment segments were lost to tech giants.

Emil Lendof/WSJ, Getty Images

By

Ben Cohen

May 1, 2026 9:00 pm ET

BPC > Only use to renew if text is incomplete or updated: | archive.is

BPC > Full article text fetched from (no need to report issue for external site): | archive.today | archive.fo

John Wall has spent nearly his entire career working for the same company. And when he tells people where he works, nobody has any clue what he’s talking about. 

“If I tell them I work at QNX,” he said, “they don’t know what that means.”

But once he explains that he technically works at the company that owns QNX, he knows exactly how they will respond: BlackBerry BB 0.14%increase; green up pointing triangle still exists? 

Yes, it does. And no, it doesn’t make phones. 

The company formerly known as Research In Motion abandoned hand-held devices a decade ago, and it feels more like a century ago when phones had clicky keyboards and everyone was obsessed with Brick Breaker.

But an astonishing number of people still rely on BlackBerry—and they don’t realize it. 

The company’s most lucrative product is not hardware but the hidden software in 275 million cars on the road today. In fact, BlackBerry’s essential technology can be found in all sorts of unexpected places, and you wouldn’t find it even if you went looking for it. 

“On a car, you’ll never see QNX’s logo,” said Wall, the division’s president. “What you will see is a better experience.” 

He likes to think of QNX engineers as plumbers and electricians, responsible for the stuff we need and never see. In a house, it’s pipes and wiring. In a car, it’s the software underpinning safety features that we take for granted. QNX is the operating system that enables all kinds of driver assistance: collision warnings, blind-spot notifications, adaptive cruise control, pedestrian detection and steering you back into a lane when you’re drifting into trouble. 

“We’re the foundation,” Wall said. “Everything pretty on top wouldn’t work without a strong foundation.” 

QNX’s operating system, which controls a number of safety features, is found in 275 million cars on the road. QNX

That foundational software has never been so valuable. As cars become computers on wheels, QNX is trusted by the world’s largest automakers because its simple, real-time operating system is designed to never, ever fail. “The only way to make this software malfunction,” a user once raved to Fortune magazine, “is to fire a bullet into the computer running it.” 

With its bulletproof reputation, the software has spread to factory floors and other workplaces that value safety, precision and tech that won’t glitch. In hospitals, for example, QNX tech is embedded in surgical robots and dozens of medical devices, which means patients are regularly putting their lives in the hands of doctors, nurses—and BlackBerry.

QNX has even become the foundation of a company left for dead long ago. 

The division that was once a rounding error is the reason that BlackBerry is suddenly making money again. Instead of being ignored by the rest of the company, QNX now accounts for half of its total revenue, and BlackBerry has strung together four straight profitable quarters for the first time since its signature product was competing with the iPhone. 

Since a bullish earnings call last month, the stock is up 50%. It’s still down 96% since its peak. But after everything the company has been through, executives are firing off a new message: BlackBerry is back. 

“The BlackBerry story,” its CEO declared on that call, “is now a growth story.” 

The story began when QNX was acquired in 2010 to help with the next generation of BlackBerrys. But by then, nothing could have helped. The company’s market cap had topped out two years earlier at $83 billion. It’s now worth $3 billion and Apple does more sales in a morning than BlackBerry does in the whole year.

Created with Highcharts 9.0.1BlackBerry's QNX FactorCars enabled by the company's QNX techSource: the company

Created with Highcharts 9.0.12018'20'250255075100125150175200225250275300 million

John Wall had a front-row seat to this spectacular crackup. 

QNX was founded in 1980 and Wall has worked for the Canadian company since he graduated college in the early 1990s. After the acquisition in 2010, many engineers moved over to RIM to build a mobile operating system for the BlackBerry. Wall stayed behind. 

While others worked on phones, his team kept plugging away at car software—and kept their QNX email addresses. 

For years, they benefited from a fantastic competitive advantage: complete and utter neglect. 

“Nobody paid attention to us,” Wall said. 

Left to their own devices, they made enough progress in car infotainment systems that Silicon Valley began paying very close attention. 

Google unveiled its own Android infotainment system, then Apple picked off QNX engineers to help build a whole car. Once again, it looked like BlackBerry was about to be squashed. 

But in 2014, with the tech giants closing in, Wall made a pivotal trip to Silicon Valley to see one of his favorite QNX customers. He also saw his future. 

Once considered obsolete, the company is now in growth mode. Ethan Miller/Getty Images

Over hefeweizens, Audi’s engineering chief told him the automaker was moving to Google for infotainment. But he wasn’t breaking up with Wall. The next generation of cars would need reliable safety features that didn’t exist yet, he explained. Instead of battling for control of the screen, Wall decided, QNX could own the software under the hood. It turned out to be the most productive beer of his life. 

“The circumstances that led to us losing infotainment pivoted the company in the right direction,” Wall said, “whether or not we knew it at the time.”

What they did know was that they didn’t have a choice. 

“There was no alternative,” he said. “We had to take what we had and figure out: Where do we go now?” 

There was only one place to go: much deeper into the car.    

After proving itself in cars, the software has found its way into medical devices, industrial automation and robotics. But going to more places hasn’t led to any more recognition for software’s plumbers and electricians. 

“The customer cares about the function of lane-keep assist,” Wall said. “They don’t care about the fact that the operating system below is blah, blah, blah.” 

Or that the blah, blah, blah is made by BlackBerry or QNX—or wherever he works.

Science of Success

More from columnist Ben Cohen about why people, teams, strategies and ideas succeed or fail

Apple Showed Me 50 Years of History That Nobody Has Ever Seen

The $40 Billion Game of Youth Sports Has Only One Winner

The Hit Movie Set Deep Inside an AI Lab That It Will Give You Goosebumps

Why There Are 29,000 People on a Waitlist for Beans

The Genius Whose Simple Invention Saved Us From Shame at the Gas Station

The Hottest Toy of the Year Is Made by a Tech Startup You’ve Never Heard Of

The Math Legend Who Just Left Academia—for an AI Startup Run by a 24-Year-Old

The Pants Cost $20. They Explain $86 Billion of Costco Sales.

Copyright ©2026 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8

Already a subscriber? Sign In

LLM (google/gemini-3.1-flash-lite-preview-20260303) summary:

• Orbital Fragility: reliance on constant human intervention creates an artificial vulnerability in low earth orbit infrastructure
• Collision Timeline: bureaucratic model predicts catastrophe within 2.8 days if automated maintenance and tracking systems fail
• Space Congestion: infinite growth of corporate megaconstellations results in objects traveling at 17000 miles per hour in overcrowded corridors
• Solar Vulnerability: geomagnetic storms create atmospheric drag that renders expensive satellite maneuver cycles unreliable and inefficient
• Operational Dependence: starlink satellites currently require active avoidance maneuvers every 1.8 minutes to prevent self-destruction
• Measurement Metric: crash clock functions as a speculative anxiety gauge for estimating the duration until debris-generating events occur
• Resource Density: artificial expansion of satellite shells creates areas with record-breaking concentrations of hardware despite obvious collision risks
• Systemic Instability: global reliance on corporate satellite networks leaves terrestrial communications at the mercy of unpredictable physical interference in space

A new study suggests that modern satellite networks may be far more fragile than they appear, with the risk of orbital collisions rising sharply if control systems are disrupted. Credit: Shutterstock

A new study warns that if satellite operators suddenly lose control during a major disruption, a catastrophic collision in orbit could happen in as little as 2.8 days.

A major solar storm does not need to smash satellites apart directly to create a crisis in orbit. It may only need to interrupt the tracking, commands, and avoidance maneuvers that keep today’s crowded satellite environment under control.

That risk is growing as low Earth orbit fills with mega constellations, large networks of satellites launched and replaced in rapid cycles. These spacecraft support internet access, communications, weather monitoring, navigation, and other services. However, they also add congestion to an orbital region where objects travel at roughly 17,000 miles per hour (27,000 kilometers per hour).

A new paper led by Sarah Thiele, who began the work as a PhD student at the University of British Columbia and is now at Princeton, attempts to measure how fragile this system has become. The study introduces a metric called the Collision Realization And Significant Harm (CRASH) Clock, which estimates how long it could take for a serious collision to occur if satellites could no longer maneuver or if operators lost reliable awareness of where objects were.

The result is alarming. Using satellite catalog data from June 2025, the researchers calculated that if operators lost the ability to send commands for avoidance maneuvers, a catastrophic collision could occur in around 2.8 days. A broader version of the CRASH Clock, based on all resident space object interactions, was 5.5 days. Back in 2018, before the rapid expansion of mega constellations, that value was 164 days.

Solar storms as a systemic threat

Satellites in low Earth orbit do not simply coast along fixed paths. They depend on station keeping, tracking updates, and collision avoidance maneuvers. According to SpaceX’s most recent biannual report cited in the study, Starlink satellites performed 144,404 collision avoidance maneuvers between December 1, 2024, and May 31, 2025. That averages 41 maneuvers per satellite per year, or one avoidance maneuver every 1.8 minutes across the Starlink network.

Paths of Starlink satellites as of Feb 2024. Credit: NASA Scientific Visualization Studio

During a major solar storm, this carefully managed system can become harder to control. Solar storms heat Earth’s upper atmosphere, causing it to expand. That increases drag on satellites, pulls spacecraft away from predicted paths, forces operators to use fuel to maintain altitude, and makes orbit forecasts less reliable.

The May 2024 “Gannon Storm” showed how disruptive this can be. Nearly half of all active satellites in low Earth orbit maneuvered because of increased atmospheric drag. The study notes that widespread repositioning, combined with unpredictable drag, made collision assessment during and after the storm much harder.

The danger grows if a storm also disrupts navigation, communications, or ground control. In that case, satellites may be harder to track just as they become less able to respond.

Why one collision matters

Kessler syndrome is the most well-known version of this kind of catastrophe, where cascading collisions fill orbit with debris and eventually make it extremely difficult to safely launch or operate spacecraft. But that runaway scenario would take years or decades to fully unfold.

To highlight the much more immediate danger, the researchers introduced a new metric called the Collision Realization and Significant Harm (CRASH) Clock, which estimates how quickly a major, debris-generating collision could become possible if active satellite control and coordination were disrupted.

Even one high-speed impact can have lasting consequences. A collision between large objects can create thousands of fragments, each becoming another hazard. Today’s debris environment is still shaped by China’s 2007 anti-satellite test involving Fengyun 1C and the 2009 collision between Iridium 33 and Kosmos 2251.

The new study finds that the densest parts of today’s satellite networks are now especially concerning. Starlink’s main shell, around 550 kilometers (342 miles) above Earth, reaches densities more than an order of magnitude higher than the peak in tracked debris near 800 kilometers (497 miles).

A shrinking margin for error

The researchers estimate that across all of low Earth orbit, close approaches within 1 kilometer (0.62 miles) occur every 36 seconds. Encounters involving at least one satellite occur about every 41 seconds, while those involving Starlink and another resident space object occur about every 47 seconds.

A close approach is not the same as a collision. Operators weigh distance, uncertainty, object size, and collision probability before deciding whether to move a satellite. Still, the frequency of these encounters shows how dependent orbit has become on fast, accurate, coordinated control.

Major solar storms are rare, but they are not hypothetical. The May 2024 Gannon Storm was the strongest geomagnetic storm in decades. The Carrington Event of September 1859 was at least twice as intense, according to the paper, and included two strong storms within a few days.

If a Carrington-scale storm occurred today, it would hit a world that relies heavily on satellites for communications, timing, Earth observation, weather forecasting, military operations, disaster response, finance, and navigation. It would also strike an orbital environment far more crowded than it was even a decade ago.

Beyond collision risk, mega constellations also contribute to debris, reentry hazards, interference with astronomy, and atmospheric pollution.

The study does not call for eliminating satellites, but it highlights a critical vulnerability. Low Earth orbit now relies on constant, precise control, and if that control is disrupted, the window to prevent a major collision could be just days.

Reference: “An orbital house of cards: Frequent megaconstellation close conjunctions” by Sarah Thiele, Skye R. Heiland, Aaron C. Boley and Samantha M. Lawler, 10 December 2025, arXiv.
DOI: 10.48550/arXiv.2512.09643

Never miss a breakthrough: Join the SciTechDaily newsletter.
Follow us on Google and Google News.

LLM (google/gemini-3.1-flash-lite-preview-20260303) summary:

• Corporate Compliance: several tech giants bowed to pentagon demands to integrate their products into classified military operations.
• Market Diversification: the defense department desperately scrambled for new vendors after labeling anthropic a supply chain risk.
• Obedient Contractors: openai google spacex microsoft amazon and nvidia signed agreements to embed their software into day to day warfare.
• Startup Integration: the pentagon included the unproven startup reflection ai in its portfolio of state sanctioned vendors.
• State Narrative: military officials claim these tools are necessary to maintain an unfair advantage and achieve absolute decision superiority.
• Open Source Ploy: the defense sector is pushing for open source models to avoid proprietary restrictions and customize tech for state use.
• Strategic Competition: administration officials framed the shift toward american controlled models as a necessary response to chinese artificial intelligence.
• Convenient Assurances: tech firms accepted vague pentagon promises that the tools would avoid mass surveillance or autonomous weapons usage.

Defense Secretary Pete Hegseth visited Capitol Hill for testimony this week. Elizabeth Frantz/Reuters

The department said Friday it is now capable of using in classified settings the technology and models from the ChatGPT maker, OpenAI; Alphabet’s GOOGL -1.22%decrease; red down pointing triangle Google; Elon Musk’s SpaceX; Microsoft; Amazon AMZN -0.76%decrease; red down pointing triangle; Nvidia NVDA 1.27%increase; green up pointing triangle; and a startup, Reflection AI. SpaceX owns Musk’s AI company, xAI.

While some of the companies, including OpenAI and SpaceX, had initial deals with the Pentagon agreeing to have their AI tools used by the military in all lawful scenarios, completing the contracts is an important step toward embedding them in day-to-day operations. The deals show how much of Silicon Valley is agreeing to the Defense Department’s terms in a way that Anthropic didn’t when it rejected the Pentagon’s contract earlier this year in what spiraled into a monthslong feud. The deals also highlight how the Pentagon is racing to incorporate AI into its systems.

For many months, Anthropic’s Claude models were some of the only tools available in classified settings through the data-mining company Palantir Technologies, which offers AI tools to the military through its platform Maven. In response to a contract dispute, the Pentagon declared Anthropic a supply-chain risk unsuitable for military work, increasing the urgency of offering other models to servicemembers in classified settings. Defense Secretary Pete Hegseth, while testifying in Congress on Thursday, called Anthropic Chief Executive Dario Amodei an “ideological lunatic.”

Emil Michael, undersecretary of defense for research and engineering and a former tech executive, said in a statement: “We are equipping the warfighter with a suite of AI tools to maintain an unfair advantage and achieve absolute decision superiority.”

Microsoft and Amazon, two of the largest providers of cloud-computing infrastructure and leading partners to model developers, already have deep relationships with the Pentagon. The two companies have their own AI tools.

The Pentagon’s deals with Nvidia and Reflection are new and highlight a desire to offer open-source models, the details of which are publicly available to developers. Most leading AI models are closed, limiting how users can customize them.

Nvidia, the world’s biggest chip company, develops its own open models. Its agreement with the Pentagon covers its Nemotron open-source models, which support AI agents capable of carrying out tasks on their own.

Nvidia Chief Executive Jensen Huang, who has cultivated ties with President Trump, has said he thinks open models are better than closed models in many national-security contexts because their attributes are fully known and they can easily be tailored for specialized use cases. “Safety and security is frankly enhanced with open-source,” he said in a recent conversation with the head of the Special Competitive Studies Project think tank that was posted online.

Nvidia Chief Executive Jensen Huang has said he supports open-source models over closed-source. Jason Henry for WSJ

Leading AI companies in China specialize in open models that they hope to sell to other countries, adding to the importance of developing U.S. alternatives, administration officials said.

Nvidia is an investor in Reflection, which specializes in open-source models and has forged close ties with the Trump administration. Led by former researchers at Google’s DeepMind lab, the company is involved in a deal supported by the government to develop models tailored for the South Korea market and is in talks to raise money from investors at a $25 billion valuation. The company hasn’t yet released any AI models.

“This shared understanding with the Pentagon is a first step in supporting U.S. national security, and sets a precedent for how AI labs could work across the U.S. government—from supporting our servicemembers to our scientists,” a Reflection spokeswoman said.

Anthropic is fighting the administration’s ban on use of its software for defense work in two separate legal cases. The company’s models have been used during the Iran war and the operation to capture then-Venezuelan President Nicolás Maduro earlier this year.

Many of the companies with newly completed agreements have said their Defense Department deals include commitments that their tools wouldn’t be used for mass surveillance or autonomous weapons. The Pentagon has said that it wouldn’t carry out those illegal activities and that companies should trust the military to use AI responsibly.

---

Watch: Anthropic CEO on the Key to Making AI ‘Safe and Controllable’

In January, Dario Amodei answered a WSJ reader’s question about potential breakthroughs in AI development in an interview with WSJ Editor-in-Chief Emma Tucker. Photo: Maurizio Martorana for The Wall Street Journal

The Global AI Race

Coverage of advancements in artificial intelligence, selected by the editors

Get WSJ's AI Newsletter

AI Has Made Memory Chips One of the World’s Most Profitable Products

OpenAI Sued by Seven Families Over Mass Shooting Suspect’s ChatGPT Use

Oracle’s Deluge of AI Debt Pushes Wall Street to the Limit

America’s Largest Landowner Is Using AI to Digitize the Forest

LLM (google/gemini-3.1-flash-lite-preview-20260303) summary:

• Staggering Financial Windfall: the reported figure of 158.4 billion dollars exists as a purely speculative accounting valuation.
• Dubious Performance Metrics: achieving the trillion dollar threshold demands unrealistic goals like an 8.5 trillion dollar corporate market capitalization.
• Shareholder Compliance Strategy: investors approved these massive payouts under the guise of maintaining executive attention at the firm.
• Speculative Future Technologies: company valuation rests on the improbable delivery of one million robotaxis and humanoid robots.
• Conditional Pay Structure: actual compensation remains tethered to hitting operational milestones rather than guaranteed simple base pay.
• Coercive Ownership Demands: the executive successfully leveraged threats of departure to increase his voting control over future development.
• Contested Equity Awards: multiple pay packages have faced significant legal scrutiny and forfeiture due to their controversial nature.
• Collateralized Share Holdings: public records indicate the chief executive maintains over 200 million company shares available for use as leverage.

Elon Musk nathan howard/Reuters

LLM (google/gemini-3.1-flash-lite-preview-20260303) summary:

• Marketing Obsession: firms are panicking to manipulate artificial intelligence responses to maintain relevance in a digital landscape.
• Engine Manipulation: executives are rebranding standard search practices into generative engine optimization to sustain the appearance of growth.
• Commodity Struggle: the company admits its core product is now a generic utility facing massive perception problems.
• Diversification Efforts: desperation to expand into contact centers and phone platforms drives the need for aggressive algorithmic positioning.
• Bureaucratic Response: leadership is forming internal swat teams to treat fundamental content shifts as if they were agile marketing campaigns.
• Buyer Avoidance: the shift in b2b behavior towards pre-sales research forces firms to infiltrate user conversations within large language models.
• Human Exploitation: the strategy relies on pressuring executives to act as corporate conduits on social media platforms to influence machine training data.
• Cultural Indoctrination: management is forcing a culture shift to ensure employees blindly prioritize the demands of automated retrieval systems.

Zoom Chief Marketing Officer Kimberly Storin is helping the company show up the right way on large language models. Zoom Communications

Yet another new job duty has skyrocketed in importance for chief marketing officers: optimizing how their companies appear in conversations with large language models like ChatGPT or Google Gemini.

For Kimberly Storin, CMO at the video meeting provider Zoom Communications, that has meant working quickly to stay on top of emerging research and trying to make sure material—whether it’s chatter on Reddit or executive commentary on LinkedIn—is showing up in a way that leads users to consider Zoom.