FBI Data Buys Reshape Privacy's Future
In 2021, Meta settled a $650 million class action under the Illinois Biometric Information Privacy Act (BIPA) for using AI to identify faces without consent.
U.S. Military's Claude LLM in Operation Absolute Resolve
An Army SBIR document and the Texas Law Review report that the U.S. military deployed Anthropic's Claude Large Language Model (LLM) in Operation Absolute Resolve in January 2026 for continuous situational awareness and targeting operations. Institutions use AI and machine learning to analyze vast datasets at unprecedented speed, transforming limited investigative tools into omnipresent digital dragnets, a point the National Endowment for Democracy made. Freedom House observed that continuous monitoring allows governments to track, identify, and predict individual behavior. The ACLU of Massachusetts found that high-resolution technologies like facial recognition and Automated License Plate Readers (LPRs) create detailed movement records, allowing police to preemptively detain activists and profile marginalized groups. LLMs automate intelligence operations by extracting deeper meaning from searched data, shifting policing logic from reactive to predictive models, a finding a peer-reviewed analysis documented. The Foundation for American Innovation reported that the commercial data broker pipeline directly contributes to this power concentration, enabling federal law enforcement to purchase detailed location histories, browsing data, app telemetry, and communication metadata on millions of Americans without a warrant. The Trump administration used databases from companies like Palantir and Babel Street to investigate civil society organizations, activists, and donors, the ACLU of Massachusetts documented.
Government Finds Social Media Monitoring Limited Value
The Brennan Center for Justice found that internal government assessments have cast doubt on the usefulness of broad social media monitoring programs, yielding "information of limited value" despite agencies continuing to collect and retain the data. The sheer volume of uncurated data, despite the concentration of power through predictive profiling, simultaneously creates analytical noise that can dilute decision-making precision and lead to arbitrary enforcement. Zdziarski highlighted that the "black box" nature of AI systems makes it difficult to explain individual determinations, obscuring how conclusions are reached. The 2023 PCLOB 702 Report states that the intelligence community does not provide metrics for the volume of incidentally collected U.S. person information under Section 702, leaving its scope unknown. This analytical noise directly leads to arbitrary enforcement and discrimination. The ACLU of Massachusetts reported that facial recognition technologies have documented misidentification issues and racial bias. The National Endowment for Democracy detailed how predictive policing programs treat Black individuals as suspects based on historical data, while content moderation algorithms disfavor Muslim speech. EPIC warned that the automation of intelligence operations creates an undue risk of automation bias, where officers respond to machine-generated prompts rather than acting as independent investigators, bypassing meaningful human oversight.
Government Surveillance Reform Act of March 2026
The Foundation for American Innovation reported that the Government Surveillance Reform Act (GSRA), introduced in March 2026, proposes requiring warrants for accessing communications, buying data from brokers, and tracking location or AI chatbot interactions. Local democratic oversight is also emerging through Community Control Over Police Surveillance (CCOPS) ordinances in cities like Boston and Cambridge, the ACLU of Massachusetts reported. These efforts, alongside new accountability structures and legislative reforms, aim to counter the asymmetry and arbitrary enforcement. The U.S. Department of State's Risk Management Profile for AI and Human Rights outlines how institutions are adopting ethical frameworks, such as the NIST AI Risk Management Framework, to ensure transparency and human rights due diligence. The Biden administration's AI executive order and an Office of Management and Budget memo established standards for federal agencies to test efficacy and mitigate bias, the Brennan Center for Justice noted. While these structures provide essential guardrails, they also impose operational costs and face limitations; for instance, the 2023 PCLOB 702 Report indicates that the Foreign Intelligence Surveillance Court (FISC) approves general procedures for Section 702 surveillance but does not review individual targeting decisions, leading to noncompliance incidents and eroded public trust.
TikTok's $92 Million Faceprint Settlement
Enzuzo reported that TikTok paid $92 million in 2021 for collecting faceprints and voiceprints, and Snap settled for $35 million in 2022 over augmented reality facial mapping. Individuals and communities are actively negotiating and maintaining privacy against continuous generative surveillance through legal challenges, community-driven data governance, and government-backed policy frameworks. In February 2025, the Ambriz, et al. v. Google LLC case successfully challenged "Google Cloud Contact Center AI" for eavesdropping on customer calls without consent, Holland & Knight detailed. The California Natural Resources Agency noted that community initiatives include Indigenous Data Sovereignty (IDSov) frameworks, such as the Yurok Tribe's Sovereign Data, Sharing, and Security Agreement in March 2026, which governs culturally sensitive information. The Cherokee Nation developed an AI policy in 2024, and the Maori Data Governance Model establishes Indigenous control over data, the California Natural Resources Agency also noted. The White House archives show that government programs and policy frameworks also support privacy negotiation; the US National Privacy Research Strategy (NPRS), updated in January 2025, prioritizes privacy-preserving system designs. The EU AI Act, with phased implementation from 2025 to 2026, introduces a risk-tiered framework prohibiting real-time biometric surveillance and requiring transparency for high-risk AI applications, Enzuzo reported.
EU AI Act's Full Implementation in 2026
As the EU AI Act's full implementation approaches in 2026, the ongoing tension between technological capabilities and fundamental freedoms will demand transparent algorithmic auditing to prevent the unchecked concentration of power. The shift to continuous generative surveillance has fundamentally altered the relationship between individuals and institutions, making privacy a continuous, proactive effort rather than a given. This erosion of default privacy, evidenced by the FBI's data purchases and the military's LLM deployments, necessitates sustained legislative action and thorough democratic oversight.
Comments ()