On August 13, 2026, Apple sent a new batch of threat notifications to users in 110 countries, and for the first time it also showed the warning on the iPhone lock screen. Apple says it has now notified users in more than 150 countries in total. It also says it "does not attribute the attacks or resulting threat notifications to any specific attackers or geographical regions." No vendor is named in the alerts, and none of the coverage we reviewed ties this wave to an Israeli company.
That absence is the policy story. Israel is where the best-known mercenary spyware originates, and its Defense Ministry is the licensor of that trade. Apple's alerts are the largest body of evidence anyone has about where the trade ends up. They currently feed nothing into Israel's licensing decisions.
The strongest case for the current regime
Israel's defenders have a real argument. Defense exports there run under the Defense Export Control Law, which took effect at the end of 2007. Its regulator, the Defense Export Controls Agency (DECA), lists human-rights considerations among its decision criteria on the Ministry of Defense's DECA page. In December 2021 the agency tightened its end-user declaration for cyber and intelligence products. Buyers must limit use to investigating and preventing serious crime and terrorism, and the declaration names sanctions for breaches, including restricting or shutting down the system, according to Defence Connect's report of the announcement.
The tightening has had measurable effects. Ctech reported that 18 Israeli companies applied for DECA cyber export permits in 2021 and only 6 in 2022, and that export markets fell from more than 100 countries to fewer than 40. An industry source told the outlet that permits were once "handed out generously" and had turned into "real stinginess". Read Ctech's account of the licensing squeeze as evidence that licensing is not a rubber stamp.
Where the system still leaks
The pressure that produced this tightening came from outside Israel. On November 4, 2021, the US Commerce Department added NSO Group and Candiru to its Entity List. It cited evidence that they "developed and supplied spyware to foreign governments that used these tools to maliciously target government officials, journalists, businesspeople, activists, academics, and embassy workers," per the Bureau of Industry and Security announcement. Israel's own regime has no comparable, publicly documented trigger that turns evidence of abuse into a revoked licence.
Amnesty International's Security Lab published "Inside Pegasus" on July 16, 2026, based on NSO documents that entered the public record in the WhatsApp and Meta case against NSO. Its most relevant passage describes how the product is licensed. Each deployment is bound by limits on the number of targets, the number of simultaneous infections and the countries that may be targeted. A customer can add countries "subject to approval by NSO Group and most likely the relevant export authority." The same report notes that certain countries, such as the United States and Israel, are barred for most customers, but that "this is a policy rather than a technical limitation." In Amnesty's analysis, controls that rest on policy have to be enforced by someone.
The design creates a gap. A licence is granted at the point of sale, and abuse surfaces years later in forensic reports, court filings and, now, platform alerts. Apple describes its notifications as high-confidence and says it withholds its detection methods so attackers cannot adapt, which is a reasonable choice. But the result is that a licensor watching the alerts learns only that mercenary spyware is active somewhere. It cannot tell whether any Israeli-licensed system is involved.
We should also be clear about what this article does not claim. Apple's alerts do not implicate Israel, NSO or any named company. The alerts are consistent with a global market that includes many vendors. The Israeli-specific accountability question is an inference about system design, not a finding about this wave.
Proportionate fixes, not a ban
Our position favours a free press, open security research and a competitive cyber sector. That rules out both extremes. Blanket bans would push capability into unlicensed jurisdictions, where there is no end-user declaration and no shutdown clause. Doing nothing leaves journalists and activists exposed to tools that Apple says cost millions of dollars per campaign and are aimed at "a very small number of specific individuals."
A narrower package would fit the evidence:
- Publish licensing statistics. DECA could report annual counts of cyber licences granted, denied and revoked, by destination category. Ctech's figures came from industry sources, not from a public ledger.
- Define a revocation trigger. The December 2021 declaration already provides for restricting or shutting down a system. A published standard would say what evidence, such as credible forensic attribution from a research lab or platform, opens a review, and who decides.
- Add independent oversight. A Knesset committee, or a retired-judge panel, should get periodic classified briefings on revocations. Executive discretion alone is not enough when the tools are used against reporters.
- Protect defensive research. Controls should be limited to intrusion systems and should not touch vulnerability research, incident response or ordinary security exports.
- Create a platform channel. Companies such as Apple could pass anonymised indicators to regulators under confidentiality rules, without revealing detection methods.
None of this needs a new statute. Most of it can be done through DECA's existing licence conditions and reporting.
What to watch
Apple's policy of not attributing alerts is defensible, since attribution errors carry costs and detection methods stay secret. It also means the pressure to close the loop has to come from licensors. Israel tightened once under US sanction threat. The better course is to build the review mechanism itself and publish how it works, before the next round of alerts arrives.