ICE Wants an AI Over Billions of Records (08/14/2026)

Two of today’s stories are about counting. ICE wants a system that can ask plain-English questions of billions of records it already holds, and the federal judiciary has decided that after nearly three decades of government hacking, somebody should write down how often a judge signs off on it. The collection came first and the count is arriving twenty-eight years late.

Top 5 Critical Privacy Alerts

1. ICE Is Shopping for an AI That Can Query Billions of Records

ICE issued a request for information on August 10 for a centralized platform that pairs generative AI with its investigative data, so agents can ask natural language questions across at least 10 terabytes covering billions of communications records. The inputs named include Title III wiretaps, pen registers, GPS pings, call detail records, cell tower data, browsing histories, financial transactions, social media from Google through WhatsApp, and forensic extractions from Cellebrite, Magnet Axiom, Oxygen Forensics, and XRY. It is scoped for at least 500 concurrent users, with responses due August 18. Biometric Update

Operator Note: Nothing here is new collection. Those feeds already exist in silos, and the silos have been doing the limiting. This is Shadow Risk at federal scale. The exposure lives in what happens the day someone joins the datasets and asks them a question in plain English.

2. Federal Courts Will Start Counting Government Spyware Use

The Administrative Office of the US Courts will begin tracking how often judges authorize spyware and hacking tools for intercepting live communications, with the numbers landing in the 2028 wiretap report published in 2029. The FBI has used these techniques since at least 1998, and there has never been a public count. For scale, Italy publishes its figures and reported spyware used against 4,321 targets in 2023. TechCrunch

Operator Note: Read the limit carefully. This covers warrants to intercept communications in real time, and it does not cover remote hacking of a device to pull data already stored on it, which runs through a different procedure. The count will be real and it will also be a floor.

3. Meta Patented Glasses That Recognize the People in Front of You

A Meta patent published today describes smart glasses that detect people in the wearer’s field of view using facial recognition, identify them, read their expressions, track the wearer’s eye gaze to find points of interest, and assemble highlight reels personalized with user relationship data. Meta previously put facial recognition code into its glasses and pulled it after press coverage, and there is no indication this ships. 404 Media

Operator Note: A patent is a statement of intent with a legal budget attached. The person whose face gets matched is not the customer here and has no interface to say no, which is the design problem underneath every wearable recognition system.

4. France’s Constitutional Council Strikes Down the Under-15 Social Media Ban

The Conseil constitutionnel struck down Article 1 of France’s law protecting minors online, which would have barred anyone under 15 from social media accounts. The ruling removes one of Europe’s more aggressive age-gating mandates and lands while other governments are building age assurance into national infrastructure. Inside Privacy

Operator Note: Every age verification mandate is a data collection mandate wearing a child safety label. You cannot prove a user is over 15 without holding something that proves how old they are, and that record then lives somewhere with a retention schedule nobody reads.

5. A Court Lets Another License Plate Reader Suit Proceed

Senior Judge William H. Orrick, in the Northern District of California, denied a motion to dismiss in McGinty v. Reimagined Parking, an Automated License Plate Recognition (ALPR) case, following the reasoning of the earlier Bartholomew decision. Two courts now agree that a claim over persistent plate reader surveillance survives the first challenge. Global Privacy Watch

Additional Privacy Alerts

Privacy Laws & Regulations

• India’s Supreme Court will examine facial recognition at protests: The challenge asks whether identifying people at demonstrations is compatible with the right to assemble. Biometric Update
• France’s data protection authority published an exploratory note on agentic AI: The Commission Nationale de l’Informatique et des Libertes (CNIL) issued it jointly with the French AI and Digital Council on July 20, and it explores the issues rather than setting binding rules. Inside Privacy
• Meta reports Australian minimum age compliance numbers: The company says it took down more than 750,000 accounts believed to belong to Australian users under 16, with Reuters reporting 462,000 deactivations. Biometric Update

Data Minimization & User Consent

• ICE agents are building a DNA database: Tens of millions of dollars have gone into collecting genetic records from millions of detainees and depositing them in a federal criminal database. PogoWasRight
• Someone hid a prompt injection in a legal filing: The instruction told any AI reading the document to side with the filer, which is what happens when a document becomes an input to a machine that acts on it. 404 Media

Privacy-Enhancing Technologies

• A new service shows you who is tracking you: Krebs walks through a tool that surfaces the trackers watching a given browser session. Krebs on Security
• NIST published new age estimation accuracy results: The takeaway is that accuracy alone does not settle whether age assurance is workable in practice. Biometric Update

Regulatory Fines & Enforcement Actions

• Scotland’s Biometrics Commissioner set out a review schedule: Eight reports and reviews are planned through autumn 2027. Biometric Update

The Axe Report is a daily briefing from Grab The Axe. Need help assessing your organization’s security posture? Take our free Human Attack Surface Score assessment.