AI Filters: Platforms Decide What You See
GSAR 552.239-7001 and Corporate Control
The U.S. General Services Administration's proposed procurement clause, "Basic Safeguarding of Artificial Intelligence Systems" (GSAR 552.239-7001), published on March 6, 2026, aims to transfer oversight from corporate architectural control to public bodies, Andrew Balch explained. Large technology corporations use their economic power, derived from user data extraction and funding of machine learning software libraries, to shape regulatory standards and often subordinate state power to corporate interests. Frontiers in Artificial Intelligence observed that content moderation practices are largely shaped by Silicon Valley elites through self-regulation, prioritizing engagement metrics over democratic transparency. Soorya Balendra's analysis for the Cambridge Forum on AI: Law and Governance revealed that AI-driven moderation decisions frequently lack context sensitivity and provide minimal explanations for content removal, leaving citizens unable to challenge or understand the algorithmic decisions affecting their information access. A study published by MediaChange indicated that this institutional pathway lowers the transaction costs of state-corporate compliance, allowing governments to outsource censorship and information control to private platforms that operate with minimal public oversight.
China Exports AI Censorship Model
A Pioneer Publisher article detailed how China exports its AI censorship model to other authoritarian regimes, fostering a global trend of AI-assisted authoritarianism. Sci-Cult documented in March 2026 that technocratic efficiency, prioritizing speed and scale, legitimizes opaque power structures and shifts accountability from visible state actors to invisible technical architectures. Freedom House noted that AI systems enable more precise censorship, preventing content from being posted at all. A Pioneer Publisher article explained that this includes predictive filters which analyze trends to preemptively throttle engagement or block discussions before they gain momentum, a practice known as "shadow banning" that moves censorship from reactive to proactive. A 2026 study from UC San Diego found that Chinese state-linked content appeared frequently in major open-source training datasets, causing commercial models prompted in Chinese to give answers more favorable to China 75.3% of the time compared to English prompts.
EU AI Act Entrenches Large Tech Firms
Major tech firms maintain their gatekeeper status through proprietary data, owning the means of algorithmic production, and controlling essential hardware and software libraries for large-scale AI training, Andrew Balch explained. He argued that despite legislative efforts like the EU AI Act, which became enforceable in 2026 and mandates transparency for generative AI, these requirements tend to entrench large tech firms as indispensable gatekeepers. Frontiers in Artificial Intelligence cautioned that an over-reliance on self-regulation by private companies has exposed individual rights and concentrated information control within Silicon Valley elites. Soorya Balendra's analysis for the Cambridge Forum on AI: Law and Governance found that Meta's AI moderation practices, for instance, face biases and underinvestment in non-English markets, particularly affecting the Global South. EPIC commented that while California's AI Transparency Act requires generative AI developers to enable content provenance, vague disclosures risk blurring the distinction between truthful and false content.
2022 Ukraine Bots Undermine Public Trust
EPIC and Brookings documented that during the 2022 invasion of Ukraine, AI-driven bot accounts comprised 60% to 80% of posts using specific pro-Russian hashtags. Algorithmically insulated information ecosystems fracture shared realities, directly weakening the domestic institutional legitimacy required for governance accountability. The New Humanitarian warned in March 2026 that social media algorithms, optimized for engagement, systematically amplify sensational and AI-generated disinformation over factual content. Brookings observed that this structural design, coupled with the "liar's dividend," degrades the shared evidentiary baseline necessary for public discourse and trust. A Gallup-Bentley University survey found that 77% of Americans distrust both businesses and government agencies to use AI responsibly, and Cornell Public Policy reported that 62% have little or no confidence in federal agencies to regulate AI effectively. According to UC San Diego and The New Humanitarian, this shift included a 4.7 percentage point increase in the likelihood of users prioritizing US Republican policy issues and a 7.4 percentage point decrease in the likelihood of viewing Ukrainian President Volodymyr Zelenskyy positively.
AI Filters Centralize Information Control
The simultaneous degradation of public verification and consolidation of corporate gatekeeping by AI filters implies a future where democratic legibility is increasingly compromised. Brookings, The New Humanitarian, and Technology Review warned that citizens face a diminished capacity for informed decision-making as the very concept of verifiable truth becomes elusive. This trajectory suggests that information control will centralize further within opaque, technocratic architectures, challenging traditional notions of democratic oversight.
Comments ()