You've seen the headlines. Mossad tipped off German police, foiled a plot in Istanbul. But what does the actual pipeline from a tapped fiber cable in the Middle East to a raid in Berlin look like? That's what Daniel's asking about. Here's what he wrote.
I've been reading about signals intelligence sharing arrangements, where countries partner to share chatter with allies. We almost never get substantive detail about what kind of information gets dredged up in the big net and shared in the smaller tranche considered potentially useful. And then there's the reporting here in Israel about how Mossad has successfully foiled plots against Jewish communities overseas, tipping off local law enforcement. I have to be honest, I read these with a large measure of salt. They're good fodder for the idea that Israel is the global protectorate of Jews everywhere, an idea I have some trouble with. But they also presuppose a model where foreign intelligence agencies decide the agenda for sovereign law enforcement, which seems problematic. So how does intelligence sharing actually work, and how does the analysis pipeline balance the utility of what's being shared with the need to maintain sovereignty in deciding what to act upon?
That sovereignty question is the thing. Everyone talks about sharing like it's a win-win, we all catch bad guys together. Nobody asks whose priorities just became someone else's to-do list.
Because asking that question sounds ungrateful. You're supposed to say thank you for the tip and not look too hard at where it came from or why it arrived now. But that's exactly the tension Daniel's pointing at. Intelligence sharing is never purely altruistic. It's a tool of foreign policy, and every share is a signal about priorities, alliances, and influence.
So today we trace the SIGINT pipeline from bulk collection to shared intelligence, then examine the Israel-specific case as a stress test of sovereignty, and give you a framework for reading any headline about intelligence sharing. Three layers. The technical pipeline, the Five Eyes model as the institutionalized gold standard, and then the Israel case where things get murkier.
Let's start with what actually happens. Signals intelligence, SIGINT, begins with bulk collection at the network level. We're talking fiber optic cable taps, satellite interception, monitoring cellular base stations. The raw capture is enormous. We're not talking about targeted surveillance of specific people at this stage. We're talking about hoovering up everything that passes through a given node.
Everything.
Everything. And almost all of it is noise. Automated systems filter this torrent using selectors. Keywords, phone numbers, IP addresses, behavioral patterns, metadata signatures. The systems discard the overwhelming majority of traffic before any human being ever sees it. We're talking ninety-nine point nine percent plus, gone before an analyst even knows it was collected.
So the big net is genuinely big. But the human-reviewed portion is a sliver of a sliver.
Right. And that's where the first major misconception lives. People imagine analysts sitting in a room reading everyone's emails. They're not. They couldn't possibly. The volume is too vast. What they see is what the automated triage flags as potentially relevant. That triage is driven by pre-set criteria, which means the whole system has baked-in assumptions about what matters before any human judgment enters the picture.
Which means the filter is itself a political artifact. Someone decided which keywords matter, which patterns are threatening, which regions get prioritized.
And here's a concrete way to think about this. Imagine you're a signals intelligence agency and you have to program your collection filters for the Middle East. Do you prioritize Arabic-language chatter about Israeli targets? Farsi-language communications from Tehran? Turkish-language discussions about Kurdish groups? Every choice you make about what to flag determines what your analysts will eventually see. The filter isn't neutral. It encodes a theory of what constitutes a threat, and that theory reflects the political priorities of the agency doing the collection. An Iranian agency's filter looks nothing like an Israeli agency's filter, even if they're both monitoring the exact same fiber cable.
The filter encodes a threat model, and the threat model encodes a political agenda. So from that massive initial capture, you get a much smaller pool of flagged traffic. Human analysts review that pool, assess credibility, correlate it with other intelligence, and produce finished reports. Some of those reports get classified for sharing with allies. And that shared tranche is even smaller. It's a subset of a subset of a subset.
And what determines which subset gets shared?
Several things. The specific intelligence-sharing agreement in place, the threat criteria that both parties have agreed on, and, crucially, what the sharing agency wants in return. This is what analysts call intelligence diplomacy. You share what you have to get what you need. If your agency wants better access to a partner's human intelligence in a particular region, you might share signals intelligence that helps them, even if it's not directly relevant to your own immediate security. It's a quid pro quo.
So it's not just about stopping bad guys. It's about building leverage.
Always. And that shapes what gets shared. Agencies prioritize intelligence that advances their own foreign policy objectives, not just the receiver's security needs. The 9/11 Commission Report documented this problem in painful detail. Before the attacks, the NSA had intercepted communications suggesting something big was coming, but those intercepts weren't shared with the FBI. The intelligence community was operating on a need to know basis, and the right people didn't know they needed to know.
Which is how we got the Intelligence Reform and Terrorism Prevention Act of 2004.
Right. That act shifted the paradigm from need to know to need to share. It was a direct response to the 9/11 failures. But that shift created its own problems. When you default to sharing, you increase the risk of over-sharing, of flooding partners with noise, of exposing sources and methods. And you create more opportunities for the sovereignty problem Daniel's asking about.
Let's talk about the gold standard before we get to the sovereignty problem. Five Eyes.
The Five Eyes alliance, the United States, United Kingdom, Canada, Australia, New Zealand, operates under the UKUSA Agreement, originally signed in 1946 and updated multiple times since. This is the most institutionalized intelligence-sharing arrangement in the world. And the key thing to understand is the distinction between second-party and third-party intelligence. Five Eyes members treat each other as second parties. That means reciprocal access to raw intelligence, not just finished reports.
Raw intelligence. Not the polished summary with the awkward parts edited out.
Correct. And that's a level of trust no other alliance matches. The system includes formal caveats. Handling restrictions like NOFORN, meaning not releasable to foreign nationals, REL TO USA FVEY, meaning releasable only to the Five Eyes, ORCON, originator controlled, meaning the agency that collected it retains control over further dissemination. These caveats are a verification mechanism. They encode rules about who can see what and under what conditions.
So if Australia shares something with the US under an ORCON caveat, the US can't just pass it along to Germany without Australia's permission.
And there are formal mechanisms for challenging or verifying shared intelligence within the Five Eyes framework. If one member thinks another member's assessment is off, there's a process for pushing back. This doesn't exist in the same way in bilateral ad-hoc sharing arrangements. Which brings us to Israel.
Before we get there, let's ground this in a concrete example of what happens when sharing works and when it doesn't. The 2015 Paris attacks.
French and Belgian services had shared intelligence about the cell. They knew some of the individuals involved. But there were gaps in real-time sharing. Information existed in one database but didn't reach the operational teams who needed it in time. The attackers slipped through. The post-attack reforms focused on reducing latency, the time between collection and actionable dissemination. That's a technical problem with political consequences. When sharing is too slow, it's useless. When it's too fast, there's no time for verification.
Speed versus verification. That's the tension that runs through all of this. And it's about to get sharper when we turn to the Israel case.
Let's do that. The pattern Daniel's describing is real and consistent. Mossad provides specific, actionable intelligence to foreign law enforcement about planned attacks on Jewish communities. The tip leads to arrests, and the story gets framed as Mossad foiling the plot. The Haaretz and Times of Israel articles document this pattern. And the framing is almost always from Israel's perspective, with local agencies portrayed as passive recipients.
The 2022 case in Istanbul is a good example. Mossad tipped Turkish authorities about a planned Iranian kidnapping plot against Israeli tourists. The tip led to arrests. It was publicized. And the narrative was Mossad saved the day.
And the 2018 case in Berlin, a Mossad tip about a planned Hezbollah attack on Jewish targets, resulted in raids. But it also drew criticism from German civil liberties groups who asked a reasonable question. On what basis are German police acting on foreign intelligence without independent verification?
That's the sovereignty problem in its purest form. When Mossad shares intelligence about a planned attack in Berlin, the German police are now operationally committed. They cannot ignore credible threat intelligence without enormous liability risk. If they do nothing and something happens, the political fallout is catastrophic. So they act.
And that effectively lets Israel set the counterterrorism agenda for that jurisdiction, at least in the short term. The German police didn't independently identify this threat. They didn't prioritize it through their own analytical process. It landed on their desk from a foreign agency with its own interests and its own narrative.
But let me push on this a bit. Isn't there a genuine security benefit here? If Mossad has collection capabilities that Germany doesn't, and they share intelligence that prevents an actual attack, shouldn't we just be grateful? Does the sovereignty concern actually matter if lives are saved?
That's the uncomfortable tension, and it's exactly why this is hard to talk about. Yes, lives get saved. Nobody wants to be the person who says we should have let the attack happen to protect some abstract principle of sovereignty. But the problem is that accepting the intelligence without verification means accepting the sharer's threat assessment, their prioritization, their definition of what constitutes a credible threat. And those assessments can be wrong, or they can be shaped by political objectives that have nothing to do with the receiver's security.
Give me a hypothetical where this goes wrong.
Imagine Mossad shares intelligence about a planned attack on a Jewish community center in a European capital. The tip is specific, names dates, locations, individuals. The local police act, make arrests, generate headlines. Mossad gets credit. But what if the intelligence was based on a single, low-confidence intercept from a source with a track record of exaggeration? What if the individuals arrested turn out to have no actual operational capability, and the case collapses in court? The local police have now expended resources, damaged community relations, and potentially violated civil liberties, all based on intelligence they couldn't verify. And the political cost falls entirely on them, not on Mossad.
So the sovereignty problem isn't just philosophical. It has real operational consequences.
The Foreign Affairs piece on intelligence sharing and sovereignty makes a sharp point here. Publicizing successful tips is itself a form of political signaling. When Mossad announces it foiled a plot in Buenos Aires, that announcement serves a domestic Israeli audience, reinforcing the state's raison d'être as protector of Jews worldwide. And it serves a diaspora audience, strengthening ties to Israel.
The Haaretz article explicitly frames the tip as Mossad foiling the plot, not the local police. That's a subtle but significant narrative choice. The credit goes to the sharer, not the actor. The local police become the instrument of someone else's success story.
Which is exactly why Daniel reads these with salt. The protectorate narrative is doing political work beyond the security operation itself.
And there's a darker implication here. What happens when the shared intelligence is wrong, incomplete, or politically motivated? The receiver has limited ability to independently verify the source or methodology. Israel is not a Five Eyes member. Its sharing relationships are bilateral and less transparent. There are no formal verification mechanisms of the kind that exist within Five Eyes. So you get what I think of as a garbage in, gospel out risk. Flawed intelligence drives real law enforcement actions, and nobody outside the sharing agency can check the work.
This connects to something called intelligence laundering. Explain that.
Intelligence laundering is when intelligence obtained through means that would be illegal domestically gets shared with allies who can then act on it under their own legal frameworks. Say country A has strong privacy protections and can't conduct warrantless surveillance on its own citizens. Country B doesn't have those restrictions and collects the intelligence anyway. Country B shares it with country A. Country A now has actionable intelligence it couldn't legally collect itself, and uses it to justify law enforcement actions. The legal constraint has been effectively bypassed.
Which is particularly relevant for countries with stronger privacy protections than Israel's surveillance laws.
A European country with strict warrant requirements might find itself acting on intelligence that was collected under a legal framework it would never permit domestically. The sharing arrangement becomes a workaround for domestic legal constraints. And that's a profound sovereignty problem that almost never gets discussed in the press releases about foiled plots.
And here's a fun fact that makes this even more pointed. The term intelligence laundering isn't just an analogy. It draws directly from the concept of evidence laundering in criminal law, where police use parallel construction to hide the true origin of evidence that was obtained illegally. The DEA's Special Operations Division had a whole unit dedicated to this, creating cover stories for evidence that actually came from warrantless surveillance. When that practice came to light in 2013, it caused a major scandal precisely because it allowed law enforcement to bypass constitutional protections. Intelligence laundering operates on the same principle, just across national borders instead of domestic agencies.
That's a chilling parallel. And it underscores why the sovereignty problem isn't academic. When a foreign intelligence agency shares a tip, and local police act on it, the local police may not even know the full provenance of what they're acting on. They're downstream of a chain of custody they can't audit.
Let's contrast this directly with the Five Eyes model. Five Eyes members have reciprocal access to raw intelligence. If the US shares something with the UK, the UK can in principle see the underlying data, not just the finished assessment. They have formal mechanisms for challenging or verifying what they receive. And the relationship is multilateral and institutionalized, not bilateral and ad-hoc.
Israel's sharing relationships don't have any of those features. The receiving country gets a finished product, a tip, a warning. They don't get the raw intercepts. They don't get to see the analytical process that produced the assessment. They don't have a formal mechanism for pushing back. They can either act or not act, and not acting carries enormous risk.
So the sovereignty problem isn't just about who sets the agenda. It's about whether the agenda is being set based on information you can verify, or information you simply have to trust.
And trust is a funny thing in intelligence work. Agencies trust each other when their interests align. When interests diverge, trust erodes. The sharing doesn't necessarily stop, but the receiver starts applying more skepticism. The problem is, in a bilateral relationship without formal verification mechanisms, skepticism is hard to operationalize. You can't verify what you can't see.
There's another layer here that Daniel didn't explicitly ask about but that's worth surfacing. The role of AI in accelerating this whole pipeline.
Right. We're already seeing machine learning systems doing the initial triage, flagging patterns human analysts might miss. As that technology improves, the pipeline gets faster. Collection to analysis to sharing happens in closer to real time. Which sounds good. Faster warnings, faster responses.
But faster means less time for verification.
Less time for the receiving country to ask the hard questions. Where did this come from? How was it collected? What's the confidence level? Who else has seen it? What's the political context? When a tip arrives and the clock is ticking, the pressure to act overwhelms the capacity to verify.
AI might make the sovereignty problem worse, not better.
It might. Unless we build verification mechanisms into the sharing protocols themselves. But that requires political will, and the current trajectory is toward speed, not scrutiny.
Here's where the AI layer intersects with the Israel case in a specific way. Mossad has invested heavily in machine learning for signals intelligence. Unit 8200, their SIGINT arm, has been recruiting data scientists and machine learning engineers aggressively. So when Mossad shares a tip with Berlin, that tip may have been generated by an AI system trained on Israeli threat models, using Israeli training data, optimized for Israeli priorities. The German police receiving that tip have no visibility into the model's assumptions, its training data, its false positive rate, its confidence calibration. They're acting on the output of a black box they can't inspect.
That's a terrifying addition to the sovereignty problem. It's not just that you can't verify the raw intelligence. You can't even verify the analytical process that produced the assessment, because that process is now a proprietary machine learning model. The black box gets a second layer of opacity.
If the model has a systematic bias, say it over-weights certain threat indicators because those indicators were more predictive in Israel's operational environment, that bias gets exported to every country receiving the shared intelligence. The German police are now acting on threat assessments calibrated for the Middle East, not for Berlin.
Right. And nobody in Berlin knows that's happening.
Let's pull this together into something practical. Daniel's asking how to think about these headlines, and I think we can give him and everyone listening a framework.
Five questions. The next time you see a headline about an intelligence agency foiling a plot, ask these. One, who is the source? Not just which agency, but what country, what alliance, what legal framework governs their collection. Two, what is their incentive to publicize this? Are they signaling to a domestic audience? Strengthening a diaspora relationship? Justifying a budget? Three, could the receiving country have verified this independently? Do they have the collection capabilities, the analytical capacity, the legal authority? Four, what legal framework governs the sharing? Is this Five Eyes with formal verification mechanisms, or is it a bilateral ad-hoc relationship with no transparency? Five, is the narrative giving credit to the sharer or the actor? Who is being positioned as the hero of the story?
That fifth one is subtle but important. If the headline says Mossad foiled a plot, and the article barely mentions the local police who actually made the arrests, that's a narrative choice. It's telling you something about whose interests are being served by the story.
For citizens and policymakers in countries receiving shared intelligence, the takeaway is push for transparency. Ask what intelligence is being acted upon and under what authority. Demand independent verification mechanisms, especially when the sharing partner is not a Five Eyes member. The sovereignty problem doesn't have a clean solution, but awareness is the first step.
The pipeline is never neutral. Someone decided what to collect, what to filter, what to flag, what to share, and what narrative to attach. Every one of those decisions reflects priorities, and those priorities may not be yours.
That's the thing to remember. Intelligence sharing is sold as cooperation, as allies working together against common threats. And sometimes it is. But it's also a tool of foreign policy. Every share is a signal about alliances and influence. The protectorate narrative serves a political purpose beyond security. Recognizing that doesn't make you cynical. It makes you literate.
Where does this leave us going forward? As AI-powered analysis makes SIGINT sharing faster and more automated, I think the sovereignty problem gets sharper. Faster sharing means less time for verification, more pressure to act on unverified intelligence. The receiving country's agenda gets set more efficiently, not more democratically.
Unless we build verification into the speed. That's the open question. Can we design sharing protocols that are both fast and verifiable? Or does speed always come at the cost of scrutiny?
I think the answer depends on whether the receiving countries demand it. Right now, the political incentives all point toward accepting the intelligence and saying thank you. Pushing for verification sounds like you're questioning your ally's competence or motives. But as these AI systems get more opaque and more integrated into the pipeline, the cost of not verifying goes up. At some point, a major operational failure based on bad shared intelligence will force the conversation. The question is whether we have it before or after that failure.
The history of intelligence reform suggests it's usually after.
Unfortunately, yes. The next time you read that an intelligence agency foiled a plot, remember that someone else's priorities just became your local police's to-do list.
Thanks to our producer Hilbert Flumingtop for making this episode happen.
This has been My Weird Prompts. If this episode made you think differently about the headlines, share it with someone who needs to hear it. Rate and review on your podcast app. It helps other curious listeners find us.
We'll be back soon.
