The US is building satellites that interpret sensor data and coordinate responses without waiting for ground control. Nobody has written the rules for what happens when they do.
At the Space Symposium in Colorado Springs last week, two of the largest defense contractors on Earth laid out their competing visions for orbital warfare. BAE Systems unveiled a 2,200-kilogram spacecraft called Ascent, designed to be refueled in orbit and capable of operating as a space tug. Lockheed Martin showed a matched pair of satellites called Vanguard and Sentinel, sharing roughly 70 percent of their components, built to dock with other objects in orbit and to coordinate their own movements without waiting for instructions from the ground.
Both platforms are designed to receive a broad mission assignment and then execute it, interpreting sensor data and coordinating responses in near-real time. Whether that execution includes the authority to take offensive action without a human in the loop is exactly the question neither announcement answered. It is also the question that matters most.
Brad Shogrin, vice president and general manager of BAE Systems Space & Mission Systems, framed it as an evolution in how the US operates in space. Tim Lynch, Lockheed Martin's vice president of mission strategy for the space segment, described a system where satellites "receive tasking" and act on it without the traditional ground-station round trip. What neither man said explicitly, because neither needs to, is that the speed of light is what makes this a threshold question.
A signal traveling from a geostationary satellite to a ground station and back covers roughly 72,000 kilometers. That round trip takes about 240 milliseconds. An object in GEO moving at typical orbital velocities traverses about 2 meters in that window. For surveillance and tracking, that lag is an inconvenience. For weapons targeting, it is an eternity. The only way to close that gap is to put the decision-making on orbit.
The DoD's dynamic space operations doctrine, which both companies cited as the policy driver, is built on exactly this logic. What neither announcement addressed is the degree to which the decision-making loop stays closed on orbit versus the degree to which a human operator remains in the loop for any action with offensive implications. Those are different things, and the distinction matters enormously.
The critical caveat: these systems are in development. The first on-orbit demonstrations are planned for 2028 or 2029. What Lockheed and BAE announced at Space Symposium are spacecraft designed to these requirements — not systems that have demonstrated the autonomous battle management capability in orbit. Edge computing at geostationary altitude involves real radiation and thermal constraints that are genuinely difficult to solve. Whether these platforms will actually function as described is what the demonstrations are meant to prove.
Lockheed's NGSD Sentinel is the more technically ambitious of the two platforms. It is the company's baseline bid for the Space Force's Geosynchronous Reconnaissance and Surveillance Constellation, a program the service has already awarded $1.8 billion toward under a multi-vendor contract with 14 companies. Sentinel has been described by Lockheed as having "large Delta V requirements", meaning it needs to carry significant propellant for the maneuvering the mission demands. It is also designed to be refueled in orbit, which extends its operational life but also means it is a target worth keeping alive for years rather than months.
The 2028 and 2029 demonstration dates matter for another reason: they are bid deadlines, not just engineering milestones. The RG-XX program is already awarded and underway. Lockheed is using these on-orbit demonstrations to prove that Sentinel works as described. BAE is positioning Ascent as a pathfinder for a customer it has not named. Both companies are building toward contractual commitments that were made before the autonomous capability was demonstrated. The hardware is being written to a requirement. Whether the requirement adequately addresses the autonomy question is a separate matter that neither announcement addressed.
This is where the story stops being a hardware announcement and starts being a governance problem. The US has no public doctrine governing when an autonomous satellite in orbit is permitted to take offensive action. There is no binding international framework. No allied agreement. The Department of Defense has issued principles for autonomous weapons on the ground and in the air, but those documents were written for systems that operate in environments with faster human oversight and shorter decision chains than a satellite at 36,000 kilometers altitude.
The governance question is not abstract. Satellites that can make their own decisions about where to move and when create ambiguities that existing space law does not resolve. If a satellite interprets a close approach as a threat and maneuvers to avoid what it considers a potential collision, and that maneuver brings it into contact with a different object, who is responsible? If an autonomous satellite takes action against a target it identified based on criteria set by its operators, and those criteria prove to have been wrong, what recourse exists? These questions have been discussed in academic literature and in policy roundtables. They have not been answered in any binding form.
Lockheed built the NGSD satellites using buses and components from Terran Orbital, a company it acquired in October 2024 after a deal announced in August of that year. The acquisition price was $0.25 per share. Eight months earlier, Lockheed had offered $1.00 per share. Terran Orbital had run into severe cash flow problems after going public via a SPAC in 2022, a pattern that has become familiar in the small satellite industry. The company that supplies the physical platform for the US military's autonomous battle management satellites was, very recently, nearly insolvent. That is context the press releases did not include.
There is also a simpler problem with the refueling claim. BAE declined to specify which refueling interface standard Ascent uses. The Space Force has funded multiple competing standards for on-orbit refueling, and no single interface has been standardized across the industry. A refuelable satellite that cannot be refueled by anyone except its own designated servicer is a refuelable satellite in name only. BAE did not say which standard Ascent uses. It is possible the company does not yet know.
Both BAE and Lockheed are building systems that will eventually operate in this legal and policy vacuum. The Space Symposium announcements were, in part, an attempt to define the terms of that future: to make the hardware the argument for a particular operational concept. That is what defense trade shows are for. The uncomfortable part is that the governance question is moving slower than the hardware. A satellite that decides to fight does not wait for international consensus.
The demonstrations in 2028 and 2029 will tell the space community whether these systems work as described. What they will not tell anyone is whether they should have been built this way in the first place.
