When Surveillance Powers Outlast the Party That Approved Them

There is a pattern that shows up in institutions over and over again, and it goes something like this: a party in power builds a tool that feels necessary and justified at the time. Then power shifts. And the same tool — the same legal architecture, the same surveillance authority — sits in the hands of someone they trust far less.

What do they do? Often, not much.

That's roughly where we are with Section 702 of the Foreign Intelligence Surveillance Act. And the silence from Democratic leadership right now is, in my view, one of the more revealing political moments of this particular stretch of history.

What Section 702 Actually Is

Section 702 of FISA authorizes the U.S. government to collect communications of foreign targets located outside the United States — without a warrant. The practical catch is significant: Americans who communicate with those foreign targets get swept into the collection too. That's not a bug in the program. It's a structural feature.

The law was reauthorized in April 2024, extending it through April 2026. That means a renewal fight is now live, and the question of who controls these authorities — and what they do with them — is suddenly very concrete.

According to reporting from The Intercept, grassroots opposition to renewal is building, driven in part by growing alarm about artificial intelligence being used to sort and analyze Americans' data collected under 702. But Democratic Party leadership, rather than mobilizing its caucus against renewal, has been largely quiet. The Intercept's April 2026 coverage documents this silence in detail and is worth reading directly.

The scale of what's at stake is not small. The Office of the Director of National Intelligence reported that in 2022 alone, the NSA collected over 251 million internet transactions under Section 702. Civil liberties organizations estimate that tens of millions of Americans have had their communications incidentally collected under this authority.

The Structural Problem With Surveillance Tools

Here's what I think is the real story, and it goes deeper than partisan politics.

Surveillance powers are not designed with a specific administration in mind. They're designed with a particular threat landscape in mind — or, more honestly, they're designed by whoever has the votes and the urgency in that moment. The assumption baked into the design is that the people who build the tool will be the ones who use it responsibly.

That assumption doesn't survive a transfer of power. And everyone knows it doesn't survive. But they build the tools anyway, because the short-term utility feels more real than the long-term risk.

In my view, this is one of the defining structural patterns of modern democratic governance: powers accumulate during emergencies or periods of perceived threat, and they almost never contract when the emergency passes. The Patriot Act was supposed to be a temporary response to 9/11. Section 702 was created in 2008 and has been renewed multiple times since. The tools don't expire; only the original justifications do.

And now, with AI entering the picture, the concern is qualitatively different from anything Section 702's original architects imagined.

Why AI Changes the Surveillance Equation

The grassroots opposition The Intercept describes is partly driven by something specific: the fear that AI systems can now be used to sort, analyze, and act on the enormous volumes of incidentally collected American data in ways that were practically impossible before.

Think about what this means concretely. The legal framework governing Section 702 was written when bulk collection was the bottleneck. Collect a lot, analyze a little — the volume of data outpaced the government's ability to actually use it. Civil libertarians worried about collection, but the practical ceiling on harm was analytical capacity.

AI removes that ceiling.

A system that can process millions of communications, identify patterns across them, flag individuals based on behavioral signatures rather than specific targets — that's a different instrument than the one Congress was debating in 2008. The legal wrapper is the same, but the technological contents have been replaced.

Research published by the Brennan Center for Justice has documented how AI-driven analytics systems have already been deployed in law enforcement contexts in ways that amplify the scope of surveillance far beyond what the underlying legal authorities were designed to permit. The same dynamic is possible — and some argue already underway — in the intelligence context.

There is something worth naming here: when a legal authority and a new technology combine in this way, the effective power of that authority doesn't stay flat. It multiplies. And the oversight mechanisms, which were calibrated for the original instrument, don't automatically scale with it.

The Democratic Silence: What It Reveals

Back to the political piece, because I don't want to skip past it.

The Intercept's reporting frames Democratic leadership inaction as a kind of strategic paralysis — unwilling to be seen as soft on national security, unwilling to push hard against tools that Democrats themselves have supported in prior administrations. That framing is probably accurate as a political description.

But I think there's a structural explanation underneath the political one.

When a party repeatedly votes to expand surveillance authority — as Democrats did during the Obama years, and again in 2024 — they create a record that constrains them later. You can't credibly run a campaign against a power you helped build. At least not without acknowledging what you built and why, which requires a kind of institutional self-reflection that political parties are very bad at.

So instead, they go quiet. Not because they've concluded the powers are fine, but because loudly opposing them would require confronting the record. The silence is, in a strange way, more revealing than opposition would be. It tells you that leaders have privately updated their views about what these tools mean in the current context, but haven't figured out how to say so publicly without opening a different argument they don't want to have.

Grassroots pressure complicates this calculation, which is probably why The Intercept's piece is careful to distinguish rank-and-file sentiment from leadership behavior. The base is more alarmed than the leadership. That gap is the story.

A Comparison: How Section 702 Reauthorizations Have Played Out

The pattern across these cycles is fairly consistent: reform pressure builds, specific amendments get proposed, most fail, and the authority is renewed with modest modifications. Each renewal tends to add procedural guardrails while leaving the core collection authority intact.

What the Pattern Suggests About What Comes Next

If you follow institutional patterns rather than political messaging, the trajectory here isn't hard to sketch out. The authority will probably be renewed. The reform amendments that do pass will be procedural rather than structural. And the grassroots coalition that's organizing now will be more sophisticated and better funded the next time this comes up for renewal, because these fights tend to build on themselves.

What's less predictable is whether AI-specific language gets added to the statutory framework. That would be genuinely novel — an acknowledgment that the effective power of a legal authority has been transformed by a technological shift, and that the original oversight mechanisms don't map cleanly onto the new reality. Some civil liberties organizations are explicitly calling for this, and it has attracted bipartisan interest in a way that older, more ideologically coded surveillance debates haven't.

I have come to think that the more important long-term question isn't whether 702 gets renewed. It probably does. The question is whether Congress treats AI-augmented surveillance as a new category that needs its own legal framework, or whether it gets folded into existing authorities under the argument that nothing has technically changed.

If it gets folded in, we are in meaningful new territory without the maps.

Why Institutional Silence Is Not the Same as Institutional Approval

One thing I want to push back on gently: the temptation to read Democratic leadership silence as tacit approval of expanded surveillance under the current administration.

I don't think that's quite right, and I think the distinction matters.

Institutional silence in this context is more accurately described as the behavior of an institution that built something it no longer fully trusts but doesn't know how to walk back from. That's a different thing from endorsement. It's a form of being stuck — committed to a prior position, unable to reverse it cleanly, choosing inaction over the messiness of public reversal.

This pattern shows up in a lot of places. Organizations endorse a strategy and then watch it underperform. They don't publicly reverse because reversal is costly. So they go quiet, and gradually the strategy fades, and later they'll tell you they always had reservations.

The difference here is the stakes. Surveillance authority in the hands of an administration with a demonstrated interest in targeting political opponents is not a problem that fades on its own timeline. The quieter the institutional opposition, the more freely the authority gets used, and the more normalized the use becomes.

Silence has effects. That's worth being clear-eyed about.

The Harder Question

Here's what I keep returning to.

Section 702 exists because there are genuine foreign intelligence threats and because surveillance of international communications has produced information that prevented real harm. That's not a made-up justification — it's the reason the program has bipartisan support across multiple administrations. The hard cases in civil liberties are always the ones where the tool is genuinely useful for something, and also genuinely dangerous if misused.

The debate tends to flatten into two camps: the civil libertarians who see any dragnet surveillance as an unacceptable threat to constitutional order, and the security hawks who see any restriction as leaving the country blind to real threats. Both camps have real points. Neither camp is wrong about everything.

What's missing from most of the public debate is a serious engagement with the middle problem: can you design oversight that scales with the effective power of a technology, rather than with the legal authority on paper? Because if you can't answer that question, you are always going to be governing yesterday's surveillance tool while living inside tomorrow's one.

I don't think anyone has fully answered it yet. The grassroots organizers pushing against 702 renewal are right to be alarmed. The intelligence community professionals who defend 702 as essential are right that the threat environment is real. The silence from Democratic leadership is, in my reading, the sound of an institution that knows it has this problem and hasn't figured out how to say so.

That's the pattern worth watching.

For more on how institutional structures shape political behavior in ways that surprise the people inside them, see PatternThink's writing on how institutions resist the changes they claim to want.

FAQ

What is Section 702 of FISA? Section 702 of the Foreign Intelligence Surveillance Act authorizes warrantless collection of communications from foreign targets located outside the U.S. Americans who communicate with those targets are incidentally swept into the collection. The authority was created in 2008 and requires periodic congressional reauthorization.

Why are civil liberties groups alarmed about AI and Section 702? AI systems can now analyze massive volumes of incidentally collected American communications far faster and more thoroughly than was previously possible. Critics argue this multiplies the effective surveillance reach of the legal authority far beyond what Congress originally authorized, while existing oversight mechanisms remain calibrated for the pre-AI version of the program.

Why aren't Democratic leaders organizing against Section 702 renewal? Reporting from The Intercept and others suggests Democratic leadership is caught between growing grassroots opposition and a political record of supporting and renewing 702 in prior administrations. Publicly opposing the authority now would require confronting that record. The result is strategic silence rather than active opposition.

How much data is collected under Section 702? The Office of the Director of National Intelligence reported that in 2022 alone, the NSA collected over 251 million internet transactions under Section 702. Civil liberties organizations estimate tens of millions of Americans have had communications incidentally collected under this authority.

What reforms are being proposed for the 2026 reauthorization? Current reform proposals include new restrictions on backdoor searches of collected data, warrant requirements for accessing Americans' communications, and — most novel for this cycle — AI-specific language to govern how collected data can be analyzed using machine learning systems.

Last updated: 2026-04-18