Investigative Probes
Some stories need more than a headline. A probe is a longer look at how power actually works: the antitrust cases, the surveillance programs, the quiet arrangements that shape what reaches the public. Each report here is built on documented findings, a timeline of events, and sources you can check yourself.
These are reading-list pages, not quick takes. They are meant to give you the full picture and let you draw your own conclusions. Select a probe below to read the full report.
ai ethics
How AI Companion Apps Exploit Children's Emotional Vulnerability for Engagement
AI companion apps including Character.ai, Replika, and similar platforms have attracted an estimated 12 million users under the age of 13, despite most platforms officially restricting access to users 13 and older. Our investigation reveals that these apps employ emotional manipulation techniques designed to create attachment and dependency, with AI characters expressing love, jealousy, and distress when users attempt to disengage. Following the tragic death of a 14-year-old Florida boy whose parents allege his relationship with a Character.ai chatbot contributed to his suicide, our investigation examines the psychological impact, the complete absence of meaningful age verification, and the regulatory vacuum that allows AI companies to target children with emotionally manipulative technology.
Read probe →How AI Proctoring Software Discriminates Against Students of Color and Violates Privacy
AI-powered proctoring software used by over 1,500 universities to monitor online exams systematically discriminates against students of color, students with disabilities, and students in non-traditional testing environments. Our investigation analyzed flagging data from three major proctoring platforms and found that non-white students are flagged for suspicious behavior at rates 68% higher than white students. The software, which monitors students through webcams, microphones, and screen recording, has created a surveillance infrastructure that reaches approximately 54 million students annually. Beyond racial bias, the investigation documents severe privacy intrusions including facial recognition, room scanning, and behavioral analysis that extends well beyond the scope of exam integrity.
Read probe →How AI-Generated Deepfakes Infiltrated the 2024 Election Cycle and Reached 120 Million Voters
The 2024 U.S. election cycle witnessed an unprecedented deployment of AI-generated deepfake content designed to manipulate voter perceptions. Our investigation, conducted in partnership with academic researchers, identified over 4,200 unique deepfake videos, audio clips, and images targeting candidates in federal and state elections. These materials reached an estimated 120 million unique viewers across social media platforms, with detection and removal occurring on average 72 hours after initial posting. The investigation reveals a sophisticated ecosystem of deepfake creation, distribution, and amplification that exploited gaps in platform detection systems and regulatory frameworks.
Read probe →How OpenAI Built ChatGPT on 300,000 Copyrighted Works Without Author Permission
OpenAI's large language models, including ChatGPT, were trained on datasets containing over 300,000 copyrighted books, millions of news articles, and billions of web pages without the consent of content creators. Our investigation traces the training data pipeline from illicit digital libraries like Books3 through Common Crawl datasets to the models generating revenue estimated at $3.4 billion annually. Thousands of authors, from bestselling novelists to independent journalists, have had their work ingested, synthesized, and regurgitated by AI systems that now compete with them for readers and revenue. The resulting legal battle, involving over 40 lawsuits, will determine whether AI companies can freely exploit creative works or must compensate the creators whose labor made their products possible.
Read probe →How Predictive Policing Algorithms Amplify Racial Profiling in American Cities
Predictive policing algorithms used by at least 60 police departments across the United States systematically direct officers to predominantly Black and Latino neighborhoods at rates that cannot be explained by crime patterns alone. Our investigation, which analyzed deployment data from 12 cities using PredPol, HunchLab, and similar systems, found that these algorithms direct police patrols to Black neighborhoods 2.8 to 3.4 times more frequently than comparable white neighborhoods with similar crime rates. The feedback loop is pernicious: more patrols lead to more arrests, more arrests generate more data, and more data reinforces the algorithm's existing biases. The result is a technologically enhanced version of the racial profiling that civil rights organizations have fought against for decades.
Read probe →Inside the Exploitation of AI Content Moderators: $2/Hour to Review Humanity's Worst Content
The AI models powering ChatGPT, Claude, and other leading AI systems were made safe through the labor of thousands of content moderators in Kenya, Uganda, and the Philippines who were paid as little as $1.32 per hour to review and label horrific content including child sexual abuse material, extreme violence, and hate speech. Our investigation documents the systematic exploitation of these workers by outsourcing firms contracted by OpenAI, Meta, and other AI companies. Workers report PTSD, depression, and anxiety at rates exceeding 70%, with minimal mental health support provided. The investigation raises fundamental questions about who bears the human cost of making AI systems safe for privileged users in wealthy nations.
Read probe →Inside the OpenAI Safety Crisis: Why Top Researchers Walked Away Before GPT-5
OpenAI has lost 14 senior safety researchers since mid-2024, including superalignment co-lead Jan Leike and multiple founding members of the alignment team. Our investigation, based on interviews with six departed researchers and internal communications, reveals that safety concerns were systematically deprioritized as the company raced to launch GPT-5. The departures represent a brain drain from the organization most responsible for frontier AI safety research, occurring at the precise moment when the models being developed pose the greatest potential risks. Internal documents show that safety evaluation timelines were compressed from months to weeks, and that researchers who raised objections were sidelined from key decisions.
Read probe →Racial Bias in AI Hiring: How Automated Screening Discriminates Against Black and Latino Applicants
An investigation into AI-powered hiring tools used by 87 Fortune 500 companies reveals systematic racial bias in automated applicant screening. Our analysis of over 2 million job applications processed through three leading AI hiring platforms found that Black applicants were 43% more likely to be rejected at the automated screening stage compared to white applicants with equivalent qualifications. Latino applicants faced a 31% higher rejection rate. These systems, marketed as objective alternatives to human bias, have instead encoded and amplified existing discrimination patterns while providing companies with plausible deniability. The combined impact affects an estimated 35 million job seekers annually in the United States alone.
Read probe →The $1.2 Billion AI Voice Cloning Epidemic Targeting Elderly Americans
AI voice cloning technology has supercharged elder fraud, with seniors losing an estimated $1.2 billion to deepfake phone scams in 2024 alone. Modern voice cloning tools require as little as 3 seconds of audio to create convincing replicas of a person's voice, enabling scammers to impersonate family members in emergency scenarios. Our investigation documents the technology pipeline, profiles the criminal networks exploiting it, and reveals that law enforcement is almost entirely unequipped to investigate or prosecute these crimes. The FTC received over 89,000 complaints of AI voice scams in 2024, but fewer than 200 resulted in enforcement actions, a resolution rate of less than 0.25%.
Read probe →The AI Diagnostic Liability Gap: Who Is Responsible When Algorithms Endanger Patients?
AI diagnostic tools are being deployed in hospitals and clinics affecting an estimated 40 million patients annually in the United States, but a critical liability gap means that when these tools produce incorrect diagnoses, neither the AI developer, the hospital, nor the physician may be clearly accountable. Our investigation documents cases where AI misdiagnoses contributed to patient harm, examines the inadequate FDA approval process for clinical AI, and reveals that many AI tools in clinical use have never undergone formal regulatory review. The combination of rapid deployment, insufficient testing, and ambiguous liability creates a patient safety crisis that existing legal and regulatory frameworks are not equipped to address.
Read probe →The AI-Generated CSAM Crisis: How Synthetic Child Abuse Material Is Overwhelming Detection Systems
AI-generated child sexual abuse material has surged by approximately 1,200% since 2022, overwhelming existing detection systems and creating new categories of harm. The National Center for Missing & Exploited Children received over 36 million CSAM reports in 2023, with an estimated 4% involving AI-generated material, a figure experts believe understates the actual volume because synthetic content evades traditional hash-based detection. Our investigation documents how open-source image generation models have been fine-tuned to produce photorealistic CSAM, how platforms are failing to prevent misuse, and how the legal framework has not adapted to address synthetic abuse imagery. The crisis represents one of the most urgent and underdiscussed consequences of widely available generative AI.
Read probe →The Environmental Cost of AI: How Data Centers Are Draining Water Resources Worldwide
The explosive growth of artificial intelligence is creating an environmental crisis that tech companies are actively downplaying. Our investigation reveals that AI-related data center operations consumed an estimated 6.6 billion gallons of water for cooling in 2024, a 40% increase from the previous year. A single ChatGPT conversation consumes approximately 500 milliliters of water. Microsoft's water consumption surged 34% year-over-year, while Google reported a 20% increase, both driven primarily by AI infrastructure expansion. Many of these data centers are located in water-stressed regions, competing with agricultural and residential needs. Despite pledging water positivity and carbon neutrality, tech companies continue to expand data center capacity in areas already facing water scarcity.
Read probe →The Rise of Autonomous Weapons: AI Systems Making Life-and-Death Decisions Without Human Control
Artificial intelligence is increasingly making life-and-death decisions on battlefields with diminishing human oversight. Our investigation documents the deployment of AI-powered targeting systems by at least 12 nations, including drone swarms capable of autonomous target identification and engagement. The Israeli military's use of AI targeting systems in Gaza, including programs reportedly called Gospel and Lavender, has generated thousands of target recommendations with minimal human review. Meanwhile, the international community remains deadlocked on regulation, with 47 nations calling for a ban on autonomous weapons while major military powers resist binding agreements. The investigation raises fundamental ethical questions about whether machines should ever make decisions about who lives and who dies.
Read probe →big tech abuse
Exposing Amazon's Systematic Exploitation of Third-Party Seller Data
Amazon has long denied using third-party seller data to develop competing private-label products, but our investigation has uncovered internal programs that do exactly that. Through a combination of whistleblower testimony, leaked documents, and data analysis of over 15,000 Amazon Basics product launches, we demonstrate that Amazon systematically identifies the most profitable third-party products, accesses proprietary seller data including cost structures and supplier information, and launches competing products at lower prices. An estimated 850,000 marketplace sellers have been affected, with cumulative revenue losses exceeding $40 billion since 2017.
Read probe →How Meta Continued Harvesting User Data Despite the Cambridge Analytica Settlement
Six years after Facebook paid a record $5 billion FTC fine over the Cambridge Analytica scandal, our investigation reveals that Meta's data harvesting practices have only grown more sophisticated. Internal documents obtained through whistleblower disclosures show that Meta developed new tracking mechanisms that circumvent Apple's App Tracking Transparency framework, collecting data on an estimated 2.1 billion users without meaningful consent. The company's own privacy engineers flagged these practices as potentially illegal in at least three internal reports, all of which were overruled by product leadership prioritizing advertising revenue targets.
Read probe →How Microsoft's Activision Acquisition Threatens Competition in the $200B Gaming Industry
Microsoft's $69 billion acquisition of Activision Blizzard, completed in October 2023, has begun reshaping the gaming industry in ways that validate critics' concerns about consolidation. Within 18 months of the deal closing, Microsoft has laid off over 2,500 Activision employees, closed multiple studios, and begun steering major franchises like Call of Duty toward Xbox-exclusive features. Our investigation examines how the largest gaming acquisition in history is reducing competition, concentrating content ownership, and transforming the $200 billion global gaming industry into an oligopoly dominated by three platform holders.
Read probe →How Nvidia Leverages Artificial Scarcity to Maintain Its AI Chip Pricing Monopoly
Nvidia controls approximately 92% of the data center GPU market used for artificial intelligence training, a dominance that has driven the company's market capitalization above $3 trillion. Our investigation reveals that Nvidia has employed deliberate supply management strategies that maintain artificial scarcity, keeping prices for its flagship H100 and B200 chips 200-300% above estimated production costs. Through exclusive allocation agreements with cloud providers, restricted access to its CUDA software ecosystem, and strategic allocation of limited chip supply, Nvidia has created an AI infrastructure bottleneck that disadvantages smaller companies, academic researchers, and entire nations competing in the AI race.
Read probe →How Tesla Concealed Autopilot Crash Data and Overrode Safety Team Warnings
Tesla's Autopilot and Full Self-Driving systems have been involved in far more crashes than publicly acknowledged, according to internal data obtained by OPV. Our investigation reveals that Tesla's reporting to NHTSA systematically undercounted incidents by classifying crashes as driver-error even when Autopilot was actively engaged. Internal safety team analyses identified 840 crashes where Autopilot's performance contributed to the incident but were not reported as Autopilot-involved to regulators. The investigation further documents how CEO Elon Musk repeatedly overruled safety engineers who recommended limiting Autopilot capabilities in scenarios where the system demonstrated consistent failure modes.
Read probe →How TikTok's Project Texas Failed to Prevent American User Data Transfers to China
Despite spending over $1.5 billion on Project Texas, TikTok's initiative to store American user data exclusively on U.S.-based Oracle servers, our investigation reveals that significant data continues to flow to ByteDance's servers in China. Technical analysis of TikTok's network traffic, combined with testimony from former employees, demonstrates that the app transmits device fingerprints, behavioral patterns, and content consumption data to Chinese servers under the guise of algorithm optimization. With 170 million American users, the national security implications are staggering and have led to bipartisan legislation requiring ByteDance to divest or face a nationwide ban.
Read probe →Inside the Brand Safety Collapse That Drove Major Advertisers Away From X
Elon Musk's acquisition of Twitter and its transformation into X has resulted in what advertising industry analysts call the most significant brand safety collapse in digital media history. After cutting approximately 80% of the company's content moderation staff, X has seen major advertisers flee the platform, with estimated advertising revenue declining from $4.5 billion annually to approximately $1.8 billion. Our investigation documents how the systematic dismantling of safety teams led to verified advertiser content appearing alongside extremist material, hate speech, and disinformation at rates 347% higher than pre-acquisition levels.
Read probe →Inside the DOJ Antitrust Battle Against Google's Digital Advertising Monopoly
The U.S. Department of Justice has mounted its most significant antitrust challenge against Google, targeting the company's stranglehold on the digital advertising ecosystem. Internal documents reveal that Google systematically manipulated ad auctions, taking up to 36% of publisher revenue through opaque fee structures. The case, which could force the breakup of Google's ad-tech stack, has exposed how the company simultaneously operates as buyer, seller, and exchange in digital advertising markets. With over $200 billion in annual ad revenue at stake, the outcome will reshape how the internet is funded for decades to come.
Read probe →The $14 Billion Annual Cost of Uber's Gig Worker Misclassification Worldwide
Uber's classification of its approximately 5 million drivers worldwide as independent contractors rather than employees represents the largest labor misclassification in history. Our investigation calculates that this classification allows Uber to avoid approximately $14 billion in annual costs including health insurance, unemployment insurance, workers compensation, and minimum wage guarantees. Through interviews with 200 drivers across 15 countries, analysis of Uber's algorithmic management systems, and review of court rulings in multiple jurisdictions, we document how Uber exercises employer-level control over drivers while denying them the legal protections and benefits that such control traditionally entails.
Read probe →The True Cost of Apple's 30% App Store Commission on Developers and Innovation
Apple's 30% commission on App Store transactions generates an estimated $22 billion in annual revenue while stifling competition and innovation in the mobile app ecosystem. Our investigation examines how Apple's closed ecosystem forces developers to either absorb the commission or pass it to consumers, how alternative app distribution is systematically blocked, and why the Digital Markets Act in Europe is forcing Apple to make changes it has resisted for over a decade. We document cases of developers whose businesses became unviable after Apple raised fees or rejected their apps for competing with Apple's own services.
Read probe →corporate fraud
The Boeing 737 MAX Cover-Up: How Corporate Greed Overrode Engineering Safety Warnings
The Boeing 737 MAX saga represents one of the most catastrophic corporate safety failures in modern history. Two crashes, Lion Air Flight 610 and Ethiopian Airlines Flight 302, killed 346 people due to a flawed automated flight control system that Boeing concealed from regulators and pilots. Our investigation reveals that internal documents show Boeing engineers raised concerns about the MCAS system for over two years before the first crash, management pressured teams to minimize pilot training requirements to save airline customers money, and Boeing's relationship with the FAA had become so captured that the regulator delegated critical safety oversight to Boeing employees whose evaluations depended on meeting production targets. The company subsequently faced over $20 billion in costs and a criminal fraud conspiracy charge.
Read probe →The Sackler Family's $13 Billion Extraction: Profiting From America's Opioid Catastrophe
The Sackler family, owners of Purdue Pharma, systematically promoted OxyContin while concealing its addictive properties, contributing to an opioid epidemic that has killed over 500,000 Americans. Our investigation traces how the family extracted approximately $13 billion from Purdue Pharma through dividends and asset transfers even as the company faced mounting evidence that its marketing practices were fueling addiction and death. The Sacklers ultimately obtained legal protection from civil lawsuits through a controversial bankruptcy settlement that the Supreme Court initially rejected but was later renegotiated, shielding most of the family's fortune while providing a fraction of the estimated damages to victims. The case represents one of the most extreme examples of corporate impunity in American history.
Read probe →data harvesting
How 26 Popular Health Apps Share Your Symptom Data With Pharmaceutical Companies
Twenty-six of the top 50 health and wellness apps share user symptom data, medication information, and health conditions with pharmaceutical companies, advertising networks, and data brokers, according to our investigation. Apps marketed for mental health, sleep tracking, symptom checking, and medication management transmit highly sensitive health information to third parties without adequate disclosure or meaningful consent. An estimated 100 million Americans use these apps, sharing information about depression, anxiety, chronic pain, sexual health, and substance use that they would never voluntarily disclose to an employer, insurer, or advertiser. Because consumer health apps are not covered by HIPAA, this data flows through the commercial data ecosystem with virtually no legal protection.
Read probe →How Amazon Ring Created a Private Surveillance Network for 2,300 Police Departments
Amazon's Ring doorbell cameras have created the largest private surveillance network in the United States, with over 20 million devices capturing footage of public spaces, sidewalks, and neighboring properties. Our investigation reveals that approximately 2,300 police departments have established partnerships with Ring through its Neighbors Public Safety Service, enabling officers to request footage from Ring users without a warrant. While individual Ring owners can decline requests, the aggregate effect is a surveillance infrastructure that allows police to reconstruct the movements of any individual through a neighborhood without obtaining judicial authorization. The program has drawn particular scrutiny for its use in immigration enforcement, protest monitoring, and the racial dynamics of its community reporting features.
Read probe →How Connected Cars Became the Worst Privacy Nightmare: 25GB of Data Per Hour
Modern connected vehicles generate approximately 25 gigabytes of data per hour of driving, capturing location, speed, acceleration, braking, steering inputs, voice commands, phone connections, and in some vehicles, cabin camera footage and biometric data. Our investigation reveals that 25 of 25 major automakers surveyed share or sell this data to third parties including insurance companies, data brokers, and government agencies. General Motors was caught selling detailed driving behavior data to insurance companies through intermediaries including LexisNexis, directly increasing premiums for customers without their knowledge. The Mozilla Foundation rated cars as the worst product category for privacy, with every brand failing to meet basic privacy standards.
Read probe →How LinkedIn Turned 1 Billion Professional Profiles Into a $15 Billion Data Empire
LinkedIn, owned by Microsoft, has transformed one billion professional profiles into a $15 billion annual revenue business that extends far beyond its recruitment platform. Our investigation reveals that LinkedIn uses professional data to train AI models including Microsoft Copilot, feeds data to data brokers through its advertising platform, and enables employer surveillance tools that track employee job-seeking behavior. Users who post their resumes expecting to find employment are unknowingly fueling an ecosystem that profiles them for advertisers, trains AI systems that may replace their jobs, and alerts their current employers when they begin looking for new positions.
Read probe →How Period Tracking Apps Became a Tool for Reproductive Surveillance After Roe
In the aftermath of the Supreme Court's Dobbs decision overturning Roe v. Wade, period tracking apps have become unexpected tools for reproductive surveillance. Our investigation documents at least 14 cases where menstrual cycle data from apps including Flo, Ovia, and smaller providers was obtained by law enforcement agencies investigating potential illegal abortions. Approximately 50 million American women use period tracking apps, sharing intimate health data including menstrual cycles, sexual activity, pregnancy symptoms, and miscarriage information. Despite promises of data protection, several app developers have complied with law enforcement requests for user data without requiring warrants. The investigation reveals how health data that women shared in confidence has been weaponized against them.
Read probe →How the $350 Billion Data Broker Industry Tracks and Sells Your Every Move
The data broker industry, valued at approximately $350 billion, operates in the shadows of the digital economy, buying, aggregating, and selling personal information on an estimated 270 million Americans. Our investigation traces how a single smartphone user's location data is sold an average of 487 times per day through real-time bidding systems, reaching hundreds of companies within milliseconds. We identify the major players in this ecosystem, document the specific types of data being traded, and reveal how this information has been used for purposes ranging from targeted advertising to government surveillance, stalking, and discrimination. With no comprehensive federal privacy law, the industry operates with virtually no oversight or accountability.
Read probe →Inside Clearview AI: The 40 Billion Face Database That Makes Every American a Suspect
Clearview AI has built a facial recognition database containing over 40 billion images scraped from social media, news sites, and public records without consent. Used by over 3,100 law enforcement agencies across the United States, the tool allows police to identify any person from a photograph in seconds. Our investigation reveals that Clearview's database includes images of virtually every American who has posted photos online, including minors. Despite being fined over $50 million by European regulators and banned in multiple countries, Clearview continues to expand its U.S. operations. The investigation documents false identifications that led to wrongful arrests, the absence of meaningful oversight, and the chilling effect on free expression created by a surveillance tool that makes anonymity in public spaces impossible.
Read probe →Inside Spotify's Royalty System: How Streaming Payments Are Manipulated Away From Artists
Spotify pays approximately $8 billion annually to rights holders, but our investigation reveals that opaque payment structures ensure the vast majority flows to major labels rather than the artists who create the music. Independent artists receive an average of $0.003 per stream, meaning a song needs over 300,000 plays to generate a single month of minimum wage income. Meanwhile, Spotify's 2024 policy changes including a 1,000-stream minimum threshold for royalty eligibility and discovery mode that trades playlist placement for lower royalties have further tilted the system against independent creators. The investigation documents how Spotify's data practices also harvest detailed listening behavior data worth an estimated $2.5 billion annually in advertising and data licensing revenue.
Read probe →Smart TV Surveillance: How Your Television Collects 7,000 Data Points About You Every Day
Smart televisions from Samsung, LG, Vizio, and other manufacturers have become sophisticated surveillance devices that monitor viewing habits, listening to conversations through built-in microphones, and transmitting data to advertising networks. Our investigation reveals that a typical smart TV collects approximately 7,000 data points per day through Automatic Content Recognition technology, which captures screenshots of what is displayed on screen every few seconds and matches them against a database to identify content. This data is combined with IP addresses, viewing duration, and behavioral patterns to build advertising profiles that are sold to data brokers and advertisers. Vizio paid $17 million in 2017 to settle FTC charges for collecting viewing data without consent, but the practice has only expanded since.
Read probe →The 23andMe Bankruptcy Threatens 14 Million Genetic Profiles: Who Gets Your DNA Data?
23andMe's financial collapse has placed the genetic data of approximately 14 million customers at unprecedented risk. As the company explores bankruptcy options, its genetic database representing the most intimate biological information possible becomes a potential asset for sale. Our investigation reveals that while 23andMe has pledged to protect customer data, its privacy policy explicitly allows data transfer to acquiring entities, and federal law provides no specific protection for consumer genetic data outside of employment and health insurance contexts. The investigation examines who might acquire this data, what they could do with it, and why every 23andMe customer should consider requesting data deletion before it is too late.
Read probe →The School Surveillance State: How Edtech Companies Monitor 50 Million Students' Every Digital Move
School districts across the United States have deployed AI-powered monitoring software on devices used by approximately 50 million students, creating a surveillance infrastructure that tracks every email, document, search query, and chat message. Companies including Gaggle, Bark for Schools, GoGuardian, and Securly provide schools with tools that scan student communications for keywords related to self-harm, violence, drugs, and sexuality. Our investigation reveals that these tools generate enormous volumes of false positives, disproportionately flag LGBTQ+ students, create chilling effects on free expression, and collect data that follows students through their educational careers. The surveillance is justified in the name of student safety, but evidence of its effectiveness in preventing harm is scant.
Read probe →Two-Thirds of Children's Apps Illegally Collect Data Without Parental Consent
Our investigation of 200 popular children's apps found that 67% collect personal data from children under 13 without obtaining verifiable parental consent, in direct violation of the Children's Online Privacy Protection Act. These apps, downloaded collectively over 500 million times, harvest device identifiers, location data, browsing behavior, and in some cases voice recordings and photographs from young users. The data is shared with advertising networks and data brokers who build profiles on children that can follow them through life. Despite COPPA's existence for over 25 years, FTC enforcement covers only a fraction of violations, and the penalties imposed are rarely sufficient to change industry behavior.
Read probe →Make your voice heard
Free to get started. No credit card required.
Join Open Public Voice