After Starlink, the West Needs Defense AI It Can Trust. Israel Can Help.

For eleven days in October, a change to Starlink service showed Ukraine how quickly a military advantage can become a dependency. Overnight on 13 October, SpaceX imposed a speed limit on Starlink terminals mounted on Ukrainian aircraft, and anything moving above about 160 kilometers an hour lost its connection. The remaining speed was significantly too slow to catch a jet-powered Shaheds. Some media reported that the order came from inside the White House, while other claimed that Elon Musk gave it himself. While the precise chain of command remains disputed, the strategic lesson is not: when essential communications depend on one outside provider, a decision made far from the battlefield can reshape operations overnight.

That concern now extends beyond satellites to artificial intelligence. As NATO incorporates Palantir’s battlefield software, the company’s European leadership has publicly insisted that it cannot remotely disable the system. Such assurances matter precisely because European governments fear becoming dependent on technology whose availability, updates or permitted uses may be decided elsewhere. Germany and France have explored alternatives, reflecting a wider desire for control rather than a simple rejection of American innovation.

The dispute between Washington and Anthropic illustrates another side of the same problem. When the Pentagon sought broader military use of Claude, the company refused to lift its limits on mass surveillance and fully autonomous targeting, and within a day the President had ordered every federal agency to stop using it and the Defense Secretary had declared an American company a supply-chain risk. Whichever side one takes, the lesson for an ally is the same. The terms on which it may fight with American AI are set in Washington and in California, by people who may change their minds in an afternoon.

Israel offers a compelling, though not perfect, answer. Decades of security pressure have pushed it to develop integrated air-defense systems in which sensors, command software and interceptors are designed to work together. Rather than purchasing a radar from one country, a missile from another and the software that connects them from a third, customers can buy a tested architecture with clearer responsibility for its operation. That is a powerful proposition for governments worried about external vetoes.

The commercial evidence is striking. Israeli defense exports reached a reported $19.2 billion in 2025, their fifth consecutive annual record, with air defense accounting for a substantial share. Germany’s Arrow 3 program demonstrates the appeal of Israeli missile-defense technology to a major European military. These purchases are not votes of confidence in a slogan; they reflect demand for systems built around an urgent and understandable mission: detecting an incoming threat, deciding how to respond and intercepting it in time.

Israel’s layered defenses also offer a useful lesson about AI. An incoming drone does not necessarily justify firing an expensive interceptor, while a ballistic missile may leave only seconds for a response. Software that combines radar data, identifies threats and helps allocate the right defensive weapon can preserve scarce ammunition and shorten response times. Laser systems such as Iron Beam could eventually make some short-range interceptions dramatically cheaper, although their effectiveness depends on conditions and they cannot replace every other layer.

Ukraine is now exploring the same principle of resilience in communications. Russian sources recently claimed that Ukrainian long-range operations were using compact SmartLite Sirius satellite terminals from Israeli company hiSky as an alternative to Starlink. A Ukrainian presidential adviser neither confirmed nor denied the claim, and cautioned that the technology is much slower than Starlink. The report therefore does not establish a wholesale switch, but it does show why alternative suppliers matter: a backup link need not match the primary system in every respect to reduce dependence on it.

The argument for Israel must nevertheless withstand scrutiny. Reporting on AI-assisted targeting has raised serious questions about human oversight, which the IDF has disputed in important respects. Microsoft also terminated a military intelligence unit’s access to certain cloud services amid concerns over surveillance. Neither controversy should be brushed aside. They underscore why customers must distinguish defensive threat-assessment tools from systems used to identify human targets, and why data governance, legal accountability and operational control belong in every procurement contract.

Europe presents a further challenge. Its SAFE financing program favors European industrial participation and imposes conditions on outside suppliers, while several governments have restricted defense trade with Israel. Israel should therefore focus on willing partners and joint production, not assume that technical excellence will override political barriers. Cooperation with countries such as Germany and Greece can combine Israeli design experience with European manufacturing, maintenance and oversight.

Nor is Israeli independence absolute. Its AI industry still relies on foreign chips, cloud infrastructure and investment, just as European and Gulf partners depend on international supply chains. Sovereignty should not mean pretending that every component can be made at home; it should mean ensuring that no single company or foreign government can disable an essential capability without alternatives. The Abraham Accords offer a potential framework for linking Israeli technology, American computing and Gulf capital, provided partners establish transparent rules for access and control.

The next generation of defense partnerships will be judged not only by accuracy, speed or price, but also by resilience under political pressure. Israel has a credible opportunity to help allies build that resilience, particularly in defensive AI and integrated air defense. The winning offer is not a promise that technology can never be switched off; it is a practical design that makes any one switch far less decisive.