Antoine Bordes, ex-head of AI at Meta FAIR now leading AI at Helsing, discusses intelligent defense, data-driven warfare, and ethical AI integration in military applications.
Intelligent defense sounds like jargon at first glance, the sort of thing you'd hear tossed around a defense tech conference between PowerPoint slides and boxed lunches.
But strip away the buzzwords, and intelligent defense is straightforward and critical.
It's the application of AI to help nations manage the overwhelming flood of information pouring from radars, satellites, drones, and countless sensors scattered across contested environments. It's about clarity in the chaos, precise decision-making in split seconds, and perhaps most important, maintaining human judgment at the center of high-stakes situations.
To unpack what intelligent defense really means, it helps to hear from someone who’s not just theorizing but actively shaping the next generation of defense technology.
Dr. Antoine Bordes is exactly that person. Formerly director of AI research at Meta's FAIR labs, Bordes spent ten formative years steering groundbreaking projects at the epicenter of AI’s explosive growth, leading al team of 600 researchers behind influential innovations like Meta’s open-source AI models, including LLaMA.
After nearly a decade at the peak of consumer-focused AI, Bordes moved into an entirely different arena. Two years ago, he joined Helsing, a fast-scaling European AI defense firm focused explicitly on applying AI to national security and defense scenarios.
He saw defense (and, he admits, healthcare) as sectors ripe for transformative change through AI and unpacked why at the Dubai AI Festival last week.
So why now for Bordes? And why defense?
AI’s recent evolution sets the stage. According to Bordes, AI's rapid expansion can be traced through distinct phases over the last 15 years. Initially, the AI movement was academic, research-oriented but around 2012-2013, deep learning emerged as the foundational architecture for most modern AI applications. “At FAIR, we were part of that foundational research–papers, university collaborations, establishing core principles," Bordes said.
Then came commercialization, the rapid industrial adoption by tech giants: Google optimizing search algorithms, Meta fine-tuning ad targeting, and Apple introducing digital assistants. “These first commercial use cases were driven by big players with vast data and resources,” he notes while the third phase is something fundamentally different.
As Bordes explains, every sector, from retail and finance to healthcare and defense, is saturated with AI initiatives. The result? Rapid experimentation, widespread use, and considerable challenges in adapting general AI methodologies to specialized (and demanding) scenarios.
This explosion of AI into those specialized sectors has changed what it means to build effective systems and for Bordes, data is at the center of every story. In defense, that narrative becomes particularly compelling.
As he outlines, wars are increasingly fought with eyes in the sky, ears underwater, and sensors scattered everywhere imaginable, something Helsing is well aware of. Bordes points to the Ukraine conflict, famously dubbed the first truly transparent war due to its unprecedented data visibility via drones, satellites, radar, and open-source intelligence flooding social media.
But, he argues, just having data isn’t enough. Defense environments require real-time interpretation and rapid decision-making. Bordes highlights electronic warfare, specifically radar detection onboard fighter jets, as a practical and vivid example of AI’s growing role.
Traditionally, radar detection was a straightforward cat-and-mouse game. Aircraft tried to avoid enemy radar signals; radars tried to detect stealthily. Today’s "agile emitters," however, are a new breed. Think adaptive radar systems continuously changing patterns and frequencies to avoid detection. Pilots need instant understanding of these shifting radar signals to avoid threats.
“Radar started in the Second World War,” Bordes told the Dubai crowd, “but now detecting radar is getting harder and harder because radars can adapt. You need algorithms embedded in planes, smart ones based on AI, that can react instantly.”
This is a prime example of intelligent defense in action: not replacing pilots, but giving them instantaneous clarity.
Building these systems comes with steep technical challenges, he says. Data quality is crucial.
“We've seen over and over that it's very hard to recover if your data is poor from the start. Even if you plan to fuse data sources hoping they correct each other’s mistakes, that rarely works as well as simply starting with clean, precise data.” Defense demands investment in better sensors, careful data selection, and precise annotation to train the algorithms effectively.
He also points to data fusion as an important area for AI-enabled defense. Systems that can combine radar, satellite, visual, and acoustic signals are the golden grail but the process of fusing those is tricky. Bordes says there is a fine art to pinpointing the highest-quality, most representative samples within massive datasets for deep annotation. This focused effort ensures that models generalize effectively to new scenarios, crucial for the unpredictable nature of defense.
He also touches on the ethical and responsible deployment of these powerful tools.
Bordes acknowledges concerns around automation and accountability but argues persuasively that greater AI involvement actually enhances ethical decision-making. Without AI, decision-makers risk information overload, leading to incomplete or delayed reactions. AI systems, in contrast, process vast amounts of real-time data rapidly, offering humans the clarity and confidence to make better-informed decisions.
“Paradoxically, putting more AI into defense situations creates more ethical outcomes,” Bordes asserts. “If humans can't quickly interpret complex real-time data streams, they're forced into partial or delayed decisions, raising the risk of error.” With AI’s help, humans remain firmly at the decision-making core, equipped with superior situational awareness.
This doesn’t mean handing full autonomy to machines. Bordes emphasizes longstanding human-machine interfaces (HMI) frameworks. Military personnel have been collaborating with automated systems for decades; Patriot missile systems, which intercept incoming missiles automatically due to split-second necessity, illustrate AI’s role clearly.
“These systems are automatic…not because humans aren't trusted, but because humans can’t physically analyze the data fast enough. Such automation is already robustly tested and documented,” he explains.
Bordes also cautions against overly restrictive or frequently changing regulations that stifle innovation. Instead, stable, long-term policies provide clarity for investment and technical development. Europe, he mentions, struggles to balance regulation against rapid technological advancement. Regulatory stability, clear guidelines, and encouraging a positive environment for AI innovation are crucial.
“What’s very hard is adapting regulations to something changing every month,” Bordes says. “Builders need courage to build. Clear regulations over a stable five-to-ten-year period provide that confidence.”
Taken together, Bordes’s insights offer an essential perspective on intelligent defense. Far from being abstract or futuristic, AI is already reshaping defense through precise algorithms, embedded intelligence, superior data fusion, and carefully managed ethical frameworks.
And, for that matter, Bordes’s transition from consumer-oriented AI at Meta to defense-focused AI at Helsing underscores just how central intelligent defense has become.