Software is eating the DoD: Brought to you by the Atlantic Council

In 2011, Marc Andreessen penned an op-ed for the Wall Street Journal proclaiming that “Software is eating the world.” Andreessen argued that every industry — even national defense — would have to embrace the “software revolution” sooner or later.

Now, Andreessen’s acolytes just have to convince the Pentagon – so long as it’s their software the department buys. Last week, the Atlantic Council launched an effort in partnership with dozens of defense industry executives — several of whom are funded by Andreessen’s firm a16z — calling on the Pentagon to usher in an era of “software-defined warfare,” a term which includes artificial intelligence and cloud computing.

In his opening remarks, the Atlantic Council’s Matthew Kroenig claimed that policymakers adopted over 70% of the recommendations from a previous commission on defense acquisition. In other words, government officials are likely taking note of the Atlantic Council’s new report.

Pointing to the threat of China, its authors argue that the Pentagon should quickly embrace software-defined warfare by increasing reliance on digital weapon testing tools, cutting software funding restrictions, and prioritizing commercial technologies. For instance, it recommends the Pentagon “enforce commercial as the default approach for software” instead of building its own custom software.

Roberto Gonzalez, a professor at San Jose State University and author of a recent paper on how Silicon Valley is transforming the military-industrial-complex, told RS in an interview that these recommendations could remove critical checks and balances; “Once you’re going full commercial and abandoning efforts in-house to produce technologies that might be of use for the Pentagon, you are putting yourself entirely at the mercy of these firms,” said Gonzalez.

The Atlantic Council’s recommendations echo those of a Pentagon commission tasked with reforming the department’s budgeting and acquisition process, known as Planning, Programming, Budgeting, and Execution (PPBE). Much like the Atlantic Council’s Commission on “software-defined warfare,” the PPBE commission was largely composed of individuals with ties to the defense industry.

Both groups recommend that the Pentagon eradicate restrictions on software funding by using research and development, procurement, or operations and maintenance funding for the “full cycle of software development, acquisition, and sustainment.” The argument is that software programs require iterative development, rendering the acquisition system’s industrial-era life-cycle stages irrelevant.

The problem is that removing funding restrictions threatens the Pentagon’s ability to evaluate program cost growth and, ultimately, its ability to terminate defective programs — many of which are dangerous and unnecessary.

"Beware of promises to speed delivery of weapons by cutting regulations,” explained Quincy Institute Senior Research Fellow William Hartung. “They may go well beyond cutting unnecessary paperwork to eliminating essential functions like independent evaluations of weapons cost and performance or guidelines to help Pentagon officials reduce price gouging.”

Additionally, the Atlantic Council Commission recommends that the Pentagon modernize test and evaluation infrastructure by simulating the “capability viability” of “digitally enabled technologies.” In other words, the Pentagon should rely on digital engineering tools and simulators to test new technologies rather than testing them in real life — which they suggest slows down the timeline for deployment while presenting operational security challenges.

Digital simulation to test and evaluate weapon systems is not new. In fact, the most expensive weapon acquisition program in U.S. history — the F-35 fighter jet, now expected to exceed $2 trillion over the course of its life cycle — can attribute some of its schedule delays to its Joint Simulation Environment (JSE).

The JSE is a digital training and testing tool designed to evaluate the F-35’s potential performance in a heavily defended airspace. The Pentagon completed testing the F-35 through the JSE in September 2023 after over seven years of development, during which time the department produced F-35s at a rate commensurate with a program already cleared through testing and evaluation.

Lockheed Martin was the primary contractor responsible for producing the F-35 while reluctantly cooperating with the Navy to develop the JSE. One can only imagine the cost growth and schedule delays possible for a software program built and tested by its own developers without any funding restrictions.

“With talk of boosting the Pentagon budget to $1 trillion per year and the procurement of complex, software-driven products on the rise, it's quite possible that more regulations may be needed to avoid shouldering the taxpayer with costly systems that do not work as advertised,” said Hartung.

Perhaps most striking about the report, however, is the fact that the authors barely mention AI. The commission is sponsored almost entirely by defense technology companies that create AI software and autonomous weapons. Gonzalez explained that if you try to sell AI to the Pentagon, some top brass may see you as a snake oil salesman. “Software, by contrast, has been around for a long time and seems less mysterious. It makes you think about floppy disks instead of autonomous swarms,” he said.

Software is much more than floppy disks. +972 Magazine reported last year that Israel employs a software known as Lavender that assigns a 1-100 score to almost every Gazan based on how likely it is they are a militant, even though it has a 10 percent false positive rate. Israeli Defense Forces then use a separate software called “Where’s Daddy?” which tracks when targets are home and marks them for bombing. In its report, the Atlantic Council makes the case for investment in “AI-enabled and software solutions” by pointing to their use by the Army’s 101st Airborne Division to address challenges with “automatic-target recognition.”

The report claims that the commissioners and consulted experts on the report participated in a “personal, not institutional, capacity.” However, the commission’s recommendations, if accepted, would pay dividends for the employers of these commissioners — the likes of which include Trae Stephens of Anduril, James Taiclet of Lockheed Martin, and Mark Valentin of Skydio. The commission is co-chaired by former Secretary of Defense Mark Esper, who is now a partner at Red Cell, a venture capital firm that invests in military start-ups such as Epirus.

The Atlantic Council's project director even closed the report launch event by telling the audience to consider investing in the commission sponsors.

“Our foundational sponsors (are) Booz Allen Hamilton, CAE, Helsing, Lockheed Martin, and Second Front Systems. As well as Aalyria, Accrete AI, Adarga, Domino Data Lab, Edge Case Research, Fathom 5, Fortem Technologies, Kodiak Robotics, Latent AI, Peraton, Primer AI, SAAB, Saronic, Scale AI, and Skydio. Phew, that was a list,” she said.

“For those looking to invest, those are companies you might want to look at,” she added.

In true DC fashion, industry-funded think tanks are happy to provide portfolio recommendations on top of national security advice — for the right price.