When the Statute Sunsets: US Spy Laws, the FISA Expiry, and the Warrantless Surveillance Regime Congress Can't Agree to Keep or Kill

On 17 April 2026, TechCrunch published a detailed account of the legislative impasse surrounding the expiry of key US surveillance authorities, reporting that lawmakers are split over protecting Americans from warrantless surveillance in ways that do not track conventional partisan lines. The story is structurally important and has received less coverage than the operational news cycle — Iran, crypto markets, AI model releases — that has dominated the technology press in April 2026. The split in Congress over surveillance reauthorisation is not a procedural footnote. It is a live question about whether the mass warrantless surveillance architecture assembled after 11 September 2001 will receive another quiet extension, be modified in ways that expand civil liberties, or — for the first time in two decades — lapse in ways that force a genuine public accounting of what that architecture actually does.

The relevant statutory provision is Section 702 of the Foreign Intelligence Surveillance Act, which authorises the National Security Agency to collect the communications of non-US persons abroad without individual warrants — a power that, as multiple court proceedings and the Snowden disclosures established, routinely captures the communications of US persons as incidental collection. The law was reauthorised in 2024 with provisions that the intelligence community characterised as safeguards and civil liberties organisations characterised as fig leaves. Those provisions are now approaching another review, and the coalition that secured the 2024 reauthorisation has not held together. The TechCrunch report documents an unusual alignment in which some conservative members of Congress, motivated by concerns about domestic surveillance of political figures, have joined civil liberties-oriented members who have opposed warrantless surveillance on constitutional grounds for decades.

The Architecture of Mass Collection

Section 702 is not a targeted surveillance programme. It is an infrastructure. The NSA collects communications data at the level of fibre trunks, data centre interconnects, and content delivery networks — the physical layer of the internet that Kate Crawford's Atlas of AI (2021) describes as the substrate of all digital activity. The legal theory under which this collection is conducted is that the agency targets foreign persons; the practical reality, documented in successive Inspector General reports and in the proceedings before the Foreign Intelligence Surveillance Court, is that the volume of US person data collected incidentally is so large that it constitutes a de facto mass domestic surveillance programme.

The framework that makes this possible is the distinction between collection and access: the government's position has consistently been that collecting data does not constitute a search, and that only querying the collected data triggers Fourth Amendment protections. Civil liberties organisations have argued for twenty years that this distinction is fictional and that the Supreme Court's jurisprudence on the third-party doctrine — which holds that information shared with a third party carries no constitutional protection — was developed in a world of paper records and cannot coherently apply to digital communications that must traverse corporate infrastructure to reach their destination. The argument has not been resolved. It has been deferred, each time, by reauthorisation.

Section 702 is not commercial surveillance. It is state surveillance conducted through commercial infrastructure. The distinction matters because the legal and political remedies available in the two cases are different. A consumer can in principle stop using a product; they cannot opt out of data collection at the fibre-trunk level. What analysis of behavioral data markets contributes is the observation that the commercial surveillance infrastructure and the state surveillance infrastructure are not separate systems — they are the same physical substrate, accessed under different legal authorities by actors whose interests sometimes align and sometimes diverge.

The Unusual Coalition and What It Reveals

The TechCrunch account emphasises the cross-party character of the opposition to clean reauthorisation. This is analytically significant because it scrambles the standard media narrative about surveillance politics: the story is not "civil libertarians versus the national security establishment." The story is more complicated, and therefore more interesting. Conservative members who were reliable supporters of post-9/11 surveillance authorities have begun raising concerns — documented in committee hearings — about the use of those authorities against political figures, including members of the current administration's predecessor. The concern is not principled opposition to warrantless surveillance; it is partisan opposition to warrantless surveillance of a specific population.

The sourcing dynamic in surveillance coverage is well established: reporters rely on authoritative sources, authoritative sources are institutional, and institutional sources have interests in what gets reported and how. The intelligence community has more credibility in mainstream reporting than civil liberties organisations; its claims about the indispensability of Section 702 are routinely taken at face value while academic and advocacy research on the programme's actual efficacy receives less prominent treatment.

What the current legislative moment has introduced is a counter-sourcing dynamic: executive branch officials who are concerned about the use of surveillance tools against political allies have begun speaking to reporters with a candour that the intelligence community's standard sourcing discipline does not usually permit. The result is a richer public picture of how the surveillance architecture actually operates — not because journalism has improved, but because the internal consensus that previously kept certain disclosures off the table has fractured along political lines.

The AI Layer Nobody Is Discussing

The surveillance reauthorisation debate is occurring at the precise moment the White House is conducting bilateral negotiations with Anthropic and other AI companies about access to their model capabilities and outputs. This conjunction is not coincidental, and it is not receiving the analytical attention it deserves. The NSA's mass collection programme was built in the 1990s and 2000s on the assumption that digital communications are the primary site of actionable intelligence. That assumption remains valid. What has changed is that the most sophisticated processing of those communications — the capacity to extract patterns, identify actors, and generate synthetic summaries of collected material — has migrated from proprietary government systems to commercial AI models that are simultaneously being deployed by intelligence agencies, being regulated (or not) by a Congress that cannot agree on their statutory authority, and being negotiated over in White House conversations that the public cannot observe.

The intelligence community's surveillance architecture is a black box: it takes in communications data and produces intelligence outputs, and the internal processes that transform inputs into outputs are not visible to the people whose data was collected. AI models are a second black box layered on top of the first: they take in the outputs of surveillance collection and produce analytical summaries, pattern matches, and recommendations, and their internal reasoning is equally opaque. The stack of black boxes — collection infrastructure, storage systems, AI processing — is the actual architecture of state surveillance in 2026, and the legislation before Congress addresses only the outermost layer.

Stakes: Expiry, Extension, and the Public Account That Has Not Been Given

Research into race and technology argues that the racialised harms of automated systems are structural, not incidental — they follow from the design choices, training data, and deployment contexts of the systems, not from the malice of individual operators. The application to mass surveillance is direct: the populations most likely to be subject to investigative follow-up based on Section 702 collection are not randomly distributed. They are disproportionately Muslim Americans, Arab Americans, and communities that have historically been subject to domestic counterintelligence operations. The cross-party coalition currently opposing clean reauthorisation is not animated by concern for those communities; it is animated by the discovery that the surveillance architecture is indiscriminate. The argument for reform that might actually protect the most exposed populations is not the argument that is moving votes.

The expiry of surveillance authorities, if it occurs without replacement, would not end the NSA's collection programme. It would create a legal gap that the intelligence community has consistently argued it cannot operate under. Whether that argument is correct — whether the intelligence community would actually halt collection rather than operate under emergency authorities or novel legal theories — is a question that the reauthorisation cycle has never been forced to answer. The current legislative impasse is the closest the question has come to being tested in two decades. The answer, when it comes, will be a primary document in the political economy of mass surveillance. The technology press, which has treated the story as secondary to AI product announcements and crypto price movements, may find it was covering the wrong numbers.

Monexus elevated the FISA expiry story above the wire's treatment because the conjunction of surveillance reauthorisation and executive AI negotiations is the structural tech story of the spring; the wire is covering them as separate beats.