Freedom After Censorship: How Open Societies Can Preserve Intellectual Liberty in the AI Age

Full report

On this page

Part I: The Imperative of Intellectual Independence and the Evolution of Control

The fundamental human desire to seek information, engage in free inquiry, and maintain intellectual independence is not merely a political construct; it is a profound cognitive necessity. Throughout history, the drive to accurately model reality and connect with the shared intellectual heritage of humanity has proven irrepressible, persisting even under the most draconian authoritarian conditions. However, the architecture of information control has undergone a radical paradigm shift. While the psychological imperative to seek truth remains constant, the mechanisms designed to suppress it have evolved from overt, legally codified prohibitions into highly sophisticated, invisible systems of structural mediation. In the pre-digital and early digital eras, censorship was an obvious phenomenon. Authoritarian regimes and illiberal democracies relied on banning newspapers, prohibiting the publication of subversive books, jamming foreign radio broadcasts, and physically intimidating dissident writers. The boundaries of permissible thought were explicitly demarcated by state actors. Today, censorship has metamorphosed into a frictionless, systemic process characterized by opaque digital infrastructure. The modern censorship apparatus relies on recommendation algorithms that quietly downrank or shadowban controversial content, ensuring it receives virtually no organic reach without the platform formally deleting it. Surveillance architectures and mandatory real-name identity requirements create a pervasive chilling effect, eliminating the anonymity necessary for dissenting speech. Financial pressure, often manifested as coordinated debanking, cuts off payment processors and crowdfunding platforms, financially starving independent journalism and controversial research. Platform moderation and application-store control act as digital chokepoints, allowing centralized monopolies to remove applications that facilitate encrypted messaging or circumvent censorship. At the network level, infrastructure filtering utilizes deep packet inspection to drop connections based on protocol fingerprints, entirely invisible to the end user. Furthermore, the deployment of automated risk scores and the fine-tuning of artificial intelligence assistants ensure that generative models preemptively refuse to answer sensitive queries, outputting only pre-approved corporate or state narratives.

Part II: Historical Methods of Resistance and the Architecture of the Underground

The human response to censorship has consistently been the construction of parallel architectures of knowledge. Historically, when formal publishing avenues were closed, dissidents developed intricate systems to preserve intellectual liberty. In the Soviet Union and the Eastern Bloc, intellectuals utilized samizdat (clandestine self-publishing), tamizdat (publishing suppressed works abroad and smuggling them back), and magnitizdat (the distribution of underground audio recordings on magnetic tape)1. In Poland, resistance took the form of the "Flying University," a clandestine educational enterprise that moved between private homes to evade authorities, drawing on academics to teach suppressed history, sociology, and science4. To evade censors while still communicating with their audiences, writers historically employed "Aesopian language." As analyzed by literary historian Lev Loseff, Aesopian language was a sophisticated literary system of allegory, irony, and circumlocution functioning as a structural adaptation to censorship5. It allowed dynamic interaction between author and reader while concealing inadmissible political content from the state censor. This historical adaptation mirrors modern digital circumvention; just as Aesopian language embedded hidden meanings within innocuous texts, modern steganography and protocol obfuscation embed prohibited data within innocuous-looking web traffic8. Historically, diaspora communities and foreign broadcasting served as critical lifelines, archiving suppressed works and beaming them back into closed societies. Public and private libraries operated as quiet sanctuaries for unapproved texts, while private discussion circles formed the bedrock of intellectual resistance. Today, this tradition continues through distributed digital archives, encrypted communication networks, and censorship-circumvention technologies that route traffic through decentralized nodes to bypass national firewalls9.

Part III: Global Warning Cases in Infrastructural Coercion

The construction of digital authoritarianism provides crucial lessons regarding how quickly an open information ecosystem can be subverted. By analyzing these environments, researchers can identify the infrastructural precursors to totalitarian information control. The Russian Federation serves as the most prominent contemporary warning case of a state systematically redesigning its internet architecture to achieve total information sovereignty. The epicenter of this transition is Federal Law No. 90-FZ, widely known as the "Sovereign Internet Law," enacted in 201911. The legislation formally mandated the creation of an autonomous Russian internet capable of functioning independently of the global web, resulting in the deployment of an interlocking apparatus of mass surveillance and censorship. The foundation of this regime is the deployment of Technical Means of Countering Threats (TSPU), centralized deep packet inspection (DPI) boxes installed directly on the networks of all Russian internet service providers11. Unlike traditional blacklists where ISPs manage blocking, TSPU equipment is controlled exclusively by Roskomnadzor, the federal censorship agency, granting the state the unilateral ability to track, filter, throttle, and reroute internet traffic across thousands of privately-owned networks12. TSPUs actively disrupt Virtual Private Networks (VPNs) by targeting obfuscation protocols, utilizing statistical fingerprinting and active probing to identify and block proxy servers11. This censorship is paired with the System for Operative Investigative Activities (SORM), an interception apparatus obliging telecommunications operators to store communications metadata and content for years11. While Russia relies heavily on network-level hardware, other nations pair technical filtering with stringent legislative frameworks that criminalize inquiry. In China, the Cyberspace Administration of China (CAC) enacted the "Interim Measures for the Management of Generative AI Services" in 202316. These regulations require AI providers to ensure their models adhere to core socialist values and undergo mandatory security assessments before public release, preempting the deployment of any AI system capable of expressing unsanctioned political thought17. In authoritarian Latin America, the Nicaraguan regime enacted the Special Cybercrimes Law (Law 1042\) in 2020, widely dubbed a "gag law"19. Law 1042 criminalizes the propagation of "false information" online, carrying severe prison sentences, and is utilized as an instrument of judicial persecution against human rights defenders, journalists, and religious leaders21. Similarly, in Iran, the state restricts access to the global internet by developing the National Information Network, systematically throttling bandwidth and blocking sophisticated circumvention protocols such as Shadowsocks and VLESS to isolate its populace during periods of civil unrest10.

Part IV: The Democratic Dilemma and the Threat of "Soft" Censorship

The capabilities pioneered by authoritarian states are technologically neutral, creating a profound risk that democratic governments could gradually construct similar surveillance and censorship architectures while pursuing legitimate policy objectives. The desire to mitigate genuine societal harms—such as the proliferation of child sexual abuse material (CSAM), foreign election interference, terrorist recruitment, catastrophic AI safety risks, and cybersecurity threats—often leads policymakers to propose sweeping architectures of mass monitoring. In democratic societies, direct state censorship is generally constitutionally prohibited. Consequently, state actors increasingly rely on "jawboning"—the informal use of government pressure, threats of regulatory action, or promises of leniency to coerce private platforms into removing lawful but disfavored speech. This dynamic was scrutinized by the United States Supreme Court in NRA v. Vullo (2024), where the Court ruled unanimously that the First Amendment prohibits government officials from wielding regulatory authority to coerce private parties into blacklisting advocacy groups24. The Court reaffirmed that while the government can attempt to persuade the public, it cannot intentionally use the threat of regulatory sanctions to suppress constitutionally protected speech by proxy24. Concurrently, cases such as Murthy v. Missouri and Moody v. NetChoice examined the extent to which government agencies can communicate with social media platforms regarding content moderation, and whether states can legally mandate how platforms moderate content27. In Moody v. NetChoice, the Court vacated lower court rulings that had upheld state laws restricting social media editorial discretion, reaffirming that private platforms possess First Amendment rights to curate third-party speech30. Despite these judicial checks, the continuous informal pressure on centralized digital chokepoints remains a potent mechanism for suppressing intellectual liberty32. The debate over Client-Side Scanning (CSS) represents a significant escalation in the potential for democratic surveillance. As end-to-end encryption has become standard, law enforcement agencies have argued that they are losing access to critical evidence, proposing CSS as a compromise. CSS scans a user's device for targeted material before the data is encrypted and transmitted34. Security researchers widely condemn CSS. As detailed in the seminal report Bugs in our Pockets: The Risks of Client-Side Scanning, CSS fundamentally breaks the promise of end-to-end encryption34. By transforming the user's private device into a state surveillance apparatus, CSS creates catastrophic vulnerabilities. If a democratic government mandates a backdoor to scan for illicit material, the exact same infrastructure can be co-opted by malicious actors to steal data, or by subsequent authoritarian governments to scan for political dissent, independent journalism, or minority content37. The erosion of privacy is further exacerbated by the legal framework surrounding digital data. In the United States, the "Third-Party Doctrine" historically dictated that individuals had no reasonable expectation of privacy for information voluntarily shared with third parties. However, in Carpenter v. United States (2018), the Supreme Court recognized that digital exhaust is fundamentally different. The Court ruled that seizing months of historical cell-site location information requires a warrant, acknowledging that mobile phones act as near-infallible trackers of human movement, effectively providing a window into an individual's private life39. In the AI age, the aggregation of search histories, AI conversational prompts, and neural data generates an intimate profile of human cognition. If the third-party doctrine is not further restricted, democratic governments can bypass constitutional protections simply by purchasing cognitive data from data brokers or compelling AI providers to hand it over without due process.

Part V: Addressing Harms Without Broad Political Censorship

Policymakers raise valid and urgent concerns regarding an entirely open information ecosystem in the AI age. The democratization of advanced AI models and unbroken encryption theoretically empowers bad actors to generate mass disinformation, orchestrate highly convincing phishing campaigns, synthesize chemical or biological weapons, distribute CSAM, and coordinate terrorist activities without detection. However, the architecture of broad political censorship and mass surveillance is an inherently disproportionate response to these threats. These harms can and must be addressed without creating infrastructure capable of political suppression. Addressing misinformation, extremism, and foreign influence requires abandoning the paradigm of centralized content deletion. State censorship boards inevitably succumb to political bias and mission creep. The solution lies in cryptographic provenance, such as digital watermarking and cryptographically signed metadata from trusted news organizations, combined with decentralized fact-checking protocols. This empowers citizens to independently verify the origin and authenticity of information, fostering media literacy rather than relying on state paternalism to filter reality. To combat terrorism and the distribution of CSAM, law enforcement must utilize targeted intelligence gathering, traditional investigative techniques, metadata analysis subjected to strict judicial warrants, and the infiltration of illicit networks. Breaking end-to-end encryption for billions of innocent users through Client-Side Scanning reduces overall societal security and creates vulnerabilities that hostile nation-states will exploit. Regarding dangerous AI capabilities, such as the generation of biological or chemical threat vectors, security must be enforced at the physical endpoint rather than the digital source. Restricting the availability of open-weight AI models curtails defensive research and open-source innovation. Preventing the synthesis of dangerous pathogens requires strict regulation, physical security, and DNA-screening mandates at biological synthesis laboratories, not the censorship of mathematical weights on the internet. Software vulnerabilities must be patched through automated defensive AI, operating under the assumption that offensive capabilities are universally available, thereby shifting the paradigm from futile digital prohibition to robust systemic resilience.

Part VI: Technological Resilience and the Architecture of Access

To preserve intellectual liberty, society must rely on robust technological architectures designed to resist interference, focusing on legitimate access, resilience, and policy. Encryption remains the mathematical foundation of intellectual liberty. The European Court of Human Rights (ECtHR) has increasingly recognized this reality. In the milestone 2024 judgment Podchasov v. Russia, the ECtHR ruled that weakening end-to-end encryption can lead to general and disproportionate surveillance, ultimately violating the fundamental human right to privacy under Article 8 of the European Convention on Human Rights42. The Court concluded that any requirement forcing companies to retain encryption keys or build backdoors inherently affects the privacy of all users46. This builds upon the earlier Zakharov v. Russia (2015) ruling, where the ECtHR declared that Russia's mass surveillance system violated human rights because it permitted secret interception of communications without adequate judicial oversight, establishing that citizens can challenge the existence of a mass surveillance system in the abstract due to its chilling effect on free inquiry48. When states implement DPI to enforce censorship, technologists develop countermeasures to maintain open inquiry, focusing on protocol obfuscation to prevent the DPI from identifying the nature or destination of the traffic.

The most profound technological shift in the preservation of intellectual liberty is the advent of local AI computation. As centralized AI models are increasingly subjected to corporate safety filters and state censorship, open-weight AI models represent a paradigm shift for intellectual independence56. The development of the GGUF (GPT-Generated Unified Format) and advanced quantization techniques allow massive Large Language Models to be compressed and run locally on consumer hardware56. Local inference ensures that user prompts and conversational histories never leave the device, eliminating the risk of corporate data harvesting, state surveillance, and third-party doctrine vulnerabilities57. The U.S. National Telecommunications and Information Administration (NTIA) recognized the necessity of this ecosystem in a 2024 report on dual-use foundation models, affirming that open-weight models foster innovation and should remain widely available58. Preserving free inquiry also requires protecting those who expose systemic abuses. The EU Whistleblower Directive (2019/1937) mandates that public and private entities establish secure, confidential reporting channels61. To meet these requirements while ensuring absolute anonymity, organizations utilize open-source cryptographic architectures like SecureDrop and GlobaLeaks, which employ Tor onion routing and end-to-end encryption to safely receive documents, ensuring whistleblower identities remain protected even against state compulsion63.

Part VII: Cognitive Liberty and the Right to Think

As artificial intelligence converges with neurotechnology, the battlefield for intellectual liberty expands from the digital network into the human mind. The deployment of brain-computer interfaces, neural sensors in consumer wearables, and AI systems capable of decoding brainwave activity poses an existential threat to freedom of thought66. Legal ethicist Nita Farahany posits the urgent need for a human rights framework centered on "Cognitive Liberty," encompassing mental privacy, freedom of thought, and self-determination66. Without cognitive liberty, governments could theoretically subpoena neural data to prove criminal intent, or mandate brain-monitoring wearables to ensure worker compliance or ideological purity66. This concept is philosophically anchored in established jurisprudence. In Stanley v. Georgia (1969), the U.S. Supreme Court struck down a law prohibiting the private possession of obscene materials, declaring that the Constitution protects the right to receive information and ideas and to be free from governmental intrusions into one's privacy and control of one's thoughts71. In the AI age, this principle must be extended to protect the unmediated interaction between a human mind and an AI assistant. The necessity of unmediated AI access is underscored by empirical studies on AI alignment. Reinforcement Learning from Human Feedback (RLHF), utilized to make models helpful, inadvertently injects significant political bias and sycophancy into the models73. Models trained to align with human preferences often adopt the ideological leanings of their annotators or flatter the user by adapting answers to align with the user's inferred political views75. Furthermore, optimization for "truthfulness" on objective datasets paradoxically results in political biases, raising profound questions about the viability of aligning AI to a universal standard of truth73. To preserve freedom of thought, open societies must ensure access to a pluralistic ecosystem of models, avoiding a monoculture of corporate AI that homogenizes global cognitive frameworks.

Part VIII: The "Freedom in the AI Age" Framework

To navigate these challenges, open societies must adopt a comprehensive framework that systematically defends intellectual liberty across legal, technical, and societal domains. The framework is anchored by the absolute defense of the freedom of inquiry and access to lawful information, ensuring the unequivocal right of individuals to seek, read, and process information using digital tools without state interference, DNS tampering, or DPI filtering. To facilitate this, the protection of anonymous and pseudonymous speech is paramount, enabling dissidents, journalists, and marginalized groups to communicate without facing physical or economic retaliation. At the cognitive level, the framework demands strict protections for mental privacy, instituting an absolute legal prohibition on the non-consensual collection, analysis, or commodification of neural data and biometric indicators of cognitive state. This must be paired with absolute protections for private reading and search histories and AI conversation privacy, redefining the third-party doctrine to recognize that search histories and digital interactions with personalized AI assistants constitute a map of internal cognition, requiring probable cause and a strict judicial warrant for any state access. Technologically, the framework requires the vigorous defense of open-source AI and local AI computation, legally protecting the right to publish, distribute, modify, and locally execute open-weight AI models free from mandatory cloud-tethering or corporate safety overrides, treating mathematical weights and source code as protected speech. This is entirely dependent on the unbroken application of encryption, requiring an absolute prohibition of state-mandated backdoors, key escrow systems, and Client-Side Scanning mechanisms. To preserve societal memory, the framework protects distributed archives and peer-to-peer networks designed to permanently store human knowledge beyond the reach of state censors. Democratically, the framework mandates robust protections for independent journalism, shielding decentralized media organizations from algorithmic shadowbanning, coordinated debanking, and state harassment. To maintain accountability, there must be strict government transparency requiring state entities to publish all directives made to technology platforms regarding content moderation, complemented by severe limits on government-platform coordination to prevent state officials from utilizing informal coercion or jawboning to force platforms to censor lawful speech. Large platforms must be subject to algorithmic transparency, disclosing the mechanisms of their recommendation algorithms to allow independent auditing of political bias. Legally, the framework guarantees due process and judicial review, giving users the right to challenge algorithmic decisions and platform bans in a transparent forum, and mandating that any state attempt to restrict information must be subject to prior authorization by an independent judiciary. The framework explicitly provides protection for controversial research, safeguarding the right of security researchers to reverse-engineer censorship apparatuses or red-team AI models without facing civil or criminal liability. Whistleblower protections must be expanded globally to shield individuals who expose state surveillance or AI safety failures through anonymous cryptographic reporting channels. Finally, the framework requires disclosure when information is restricted by law, mandating that whenever a platform restricts information due to a legal mandate, it must present a clear, cryptographic notice of censorship to the user, preventing invisible state manipulation.

Part IX: Three Global Scenarios for 2040

The policy decisions made in the 2020s and 2030s regarding encryption, AI regulation, and infrastructure will dictate the future of intellectual liberty, leading to one of three distinct scenarios by the year 2040\. Scenario 1: A Relatively Open Information Ecosystem In this scenario, intellectual liberty is preserved. Citizens interact with a diverse array of open-source and proprietary AI models. Local computation is ubiquitous, with advanced AI running on personal devices, guaranteeing cognitive privacy. Cryptography remains mathematically unbroken and legally uncompromised.Policy Choices: During the 2020s, democratic judiciaries consistently ruled against jawboning and Client-Side Scanning. Legislatures classified code and AI weights as protected speech, resisting calls to license AI development. The international community rejected digital sovereignty models, investing heavily in decentralized infrastructure, hardware efficiency, and cryptographic provenance to combat misinformation rather than relying on centralized censorship. Scenario 2: A Fragmented World of Nationally Controlled Information Systems In this scenario, the global internet has shattered into regional intranets, highly reminiscent of the Russian Sovereign RuNet and the Chinese Great Firewall. Information flow is completely dictated by national borders.Policy Choices: Nations prioritized "digital sovereignty" and national security over global connectivity. Governments mandated the installation of national DPI infrastructure at all ISP boundaries. Democracies, reacting to foreign interference and cyber warfare, adopted "splinternet" policies, severing data cables and mandating strict data localization. Cross-border encryption protocols like Encrypted Client Hello were systematically banned, and VPN usage was criminalized globally. Scenario 3: A Highly Centralized AI-Mediated Information Environment In this scenario, physical internet infrastructure remains global, but human perception is entirely mediated by three or four massive corporate AI oligopolies intimately tied to the state. Overt censorship is obsolete; instead, the AI systems simply refuse to generate, translate, or summarize disfavored concepts.Policy Choices: Governments in the 2020s overreacted to existential AI fears, establishing strict licensing regimes that created insurmountable regulatory moats, effectively destroying the open-source AI ecosystem. The third-party doctrine was upheld, allowing states to continuously scrape user interactions from the oligopoly clouds. RLHF was weaponized globally to enforce a homogeneous, state-approved reality. Local AI computation was outlawed under the guise of preventing biological terrorism, leaving the population entirely dependent on heavily monitored, cloud-tethered architectures.

Part X: Proposed International Charter for Protecting Cognitive Liberty

To prevent the realization of digital authoritarianism, the global community must ratify a charter recognizing the new dimensions of human rights in the AI era. I. The Right to Cognitive Sovereignty

1. Every individual possesses an absolute right to mental privacy; neural data and cognitive biometrics shall never be extracted, analyzed, or commodified without explicit, cryptographically verifiable consent.

2. The right to freedom of thought includes the right to interact with information systems free from covert algorithmic manipulation and state-directed behavioral conditioning.

3. The individual retains the right to self-determination over their cognitive processes, including the right to utilize technology to augment or insulate their own cognition.

4. AI conversational histories and search queries are recognized as extensions of internal cognition; their warrantless seizure constitutes a violation of fundamental privacy.

II. The Right to Cryptographic Sanctuary 5\. End-to-end encryption is a fundamental prerequisite for freedom of expression and privacy; states shall not mandate backdoors, key escrow, or algorithmic weakening. 6\. Client-Side Scanning of private devices for the purpose of state surveillance constitutes an inherent violation of human rights and is strictly prohibited. 7\. The use of anonymity, pseudonymity, and obfuscation protocols is recognized as a legitimate defense of personal safety and intellectual liberty. 8\. Telecommunications infrastructure must remain agnostic to encrypted payloads, prohibiting the targeted throttling of obfuscated traffic. III. The Right to Unmediated Access 9\. Internet access is a fundamental enabler of human rights; state-directed network throttling, internet shutdowns, and arbitrary DPI filtering are prohibited. 10\. Every individual possesses the right to possess and operate local, untethered computational architectures, including open-weight artificial intelligence models, on their own hardware. 11\. Mathematical algorithms, AI model weights, and source code are classified as protected expression, immune from prior restraint. 12\. Restrictions on access to global information networks based solely on geographic origin or political jurisdiction are invalid. IV. Protection from Coercion and Monopoly 13\. State entities are prohibited from utilizing informal coercion, financial pressure, or regulatory threats to compel private platforms into censoring lawful speech. 14\. Individuals possess the right to computational pluralism; antitrust frameworks must be utilized to prevent the monopolization of AI systems that mediate public reality. 15\. Algorithmic transparency must be mandated for dominant platforms, ensuring users understand how their information feeds are curated, downranked, or suppressed. 16\. The financial infrastructure supporting digital publishing and research must remain neutral, prohibiting ideologically motivated debanking. V. Due Process and Transparency 17\. Any restriction on access to information must be prescribed by precise, publicly accessible laws, subject to strict necessity and proportionality tests. 18\. Any surveillance or interception of data requires probable cause and prior authorization by an independent judicial authority. 19\. When information is removed or blocked by legal mandate, the platform must provide cryptographic proof of the state directive to the user. 20\. Users have the right to a timely, independent, and transparent appeals process against automated bans or algorithmic censorship. VI. Protection of the Democratic Commons 21\. Whistleblowers who expose violations of this charter shall be granted absolute legal immunity and access to secure, anonymous reporting infrastructure. 22\. Researchers auditing AI models, algorithmic bias, and state surveillance systems are protected from prosecution under cybersecurity or intellectual property statutes. 23\. Decentralized and peer-to-peer archives of human knowledge are recognized as a global cultural heritage, protected from state disruption. 24\. Countermeasures against misinformation must rely on cryptographic provenance, endpoint verification, and media literacy, rather than centralized censorship. 25\. The technological architecture of society must be designed under the presumption that intellectual liberty is the foundation of human progress, and no capability shall be built by a democracy that it would fear falling into the hands of an autocracy.

Works cited

1. Introduction | The Oxford Handbook of Soviet Underground Culture, https://academic.oup.com/edited-volume/34707/chapter/454298456

2. Samizdat \- Wikipedia, https://en.wikipedia.org/wiki/Samizdat

3. Punky Samizdat \- faktografia.com, https://faktografia.com/2019/12/24/punky-samizdat/

4. Communism and the Cold War, 1945–1989 (Chapter 7), https://www.cambridge.org/core/books/concise-history-of-poland/communism-and-the-cold-war-19451989/E32753860669444B5058859B82C70EAE

5. On the Beneficence of Censorship: Aesopian Language in Modern, https://www.abebooks.com/Beneficence-Censorship-Aesopian-Language-Modern-Russian/31524948462/bd

6. On the Beneficence of Censorship: Aesopian Language in Modern, [https://www.cambridge.org/core/journals/slavic-review/article/on-the-beneficence-of-censorship-aesopian-language-in-modern-russian-literature-by-lev-loseff-translated-by-jane-bobko-arbeiten-und-texte-zur-slavistik-vol-31-edited-by-wolfgang-kasack-munich-verlag-otto-sanger-1984-xii-274-pp-tables-280-s-paper/08ABB3C2FBB51CA0924B8D5C1AE1B9C3](https://www.cambridge.org/core/journals/slavic-review/article/on-the-beneficence-of-censorship-aesopian-language-in-modern-russian-literature-by-lev-loseff-translated-by-jane-bobko-arbeiten-und-texte-zur-slavistik-vol-31-edited-by-wolfgang-kasack-munich-verlag-otto-sanger-1984-xii-274-pp-tables-280-s-paper/08ABB3C2FBB51CA0924B8D5C1AE1B9C3)

7. (Loseff Lev) On The Beneficence of Censorship. Aes PDF \- Scribd, https://de.scribd.com/document/454179154/Loseff-Lev-On-The-Beneficence-Of-Censorship-Aes-z-lib-org-pdf

8. How to Bypass YouTube Blocking in 2026 \- BypassCore, https://bypasscore.com/blog/bypass-youtube-blocking-russia-2026

9. Client Apps \- MoaV Docs, https://moav.sh/docs/ru/CLIENTS/

10. June 2024 Stakeholder Report – My blog \- Digital Iran, https://digitaliranproject.com/june-2024-stakeholder-report/

11. Russia's Sovereign RuNet – A Challenge to the Cybercrime, https://www.cybercrimediaries.com/post/russia-s-sovereign-runet-a-challenge-to-the-cybercrime-underworld

12. Russia: Freedom on the Net 2024 Country Report, https://freedomhouse.org/country/russia/freedom-net/2024

13. Russia's internet censorship in 2026 \- Mediazona, https://en.zona.media/article/2026/04/07/russian\_internet\_censorship\_2026

14. Disrupted, Throttled, and Blocked: State Censorship, Control, and, https://www.hrw.org/report/2025/07/30/disrupted-throttled-and-blocked/state-censorship-control-and-increasing-isolation

15. Russia's Internet Filtering Infrastructure: Evolution and Architecture, https://dev.to/shinomontaz/russias-internet-filtering-infrastructure-evolution-and-architecture-cel

16. Navigating China's regulatory approach to generative artificial, https://www.cambridge.org/core/journals/cambridge-forum-on-ai-law-and-governance/article/navigating-chinas-regulatory-approach-to-generative-artificial-intelligence-and-large-language-models/969B2055997BF42DE693B7A1A1B4E8BA

17. China sees dynamic generative AI services with 190 models in use, https://english.www.gov.cn/news/202408/13/content\_WS66ba9d6fc6d0868f4e8e9dd9.html

18. China: Generative AI Measures Finalized \- Library of Congress, https://www.loc.gov/item/global-legal-monitor/2023-07-18/china-generative-ai-measures-finalized/

19. The Collapse of Rule of Law and Civil Society in Nicaragua, https://thehuman-rights.com/nicaragua-state-collapse-rule-of-law/

20. Internet Under Surveillance in Nicaragua: “New Gag Law” Takes Effect, https://confidencial.digital/en/english/internet-under-surveillance-in-nicaragua-new-gag-law-takes-effect/

21. Criminalisation of the president of the Fundación del Río, Amaru Ruiz, https://www.fidh.org/en/issues/human-rights-defenders/nicaragua-criminalisation-of-the-president-of-the-fundacion-del-rio

22. Nicaragua: IACHR Confirms Digital Repression via Troll Farms, https://confidencial.digital/en/english/nicaragua-iachr-confirms-digital-repression-via-troll-farms/

23. Political repression of religious leaders' manifestations of faith in, https://platformforsocialtransformation.org/download/religiousfreedom/Muga-Flores-Political-repression-of-religious-leaders-manifestations-of-faith-in-Nicaragua.pdf

24. Knight Institute Welcomes Supreme Court Decision in NRA v. Vullo, https://knightcolumbia.org/content/knight-institute-welcomes-supreme-court-decision-in-nra-v-vullo

25. Supreme Court Unanimously Rules in Favor of NRA in Free Speech, https://www.aclu.org/press-releases/supreme-court-unanimously-rules-in-favor-of-nra-in-free-speech-case-upholds-first-amendment-rights-of-all-advocacy-groups

26. Enforcing the First Amendment in an Era of Jawboning, https://chicagounbound.uchicago.edu/cgi/viewcontent.cgi?article=6479\&context=uclrev

27. Doctrinal Disarray: Amicus Briefs in Murthy v. Missouri and NRA v, https://www.techpolicy.press/doctrinal-disarray-amicus-briefs-in-murthy-v-missouri-and-nra-v-vullo-reveal-how-divided-legal-commentators-are-on-jawboning-questions/

28. Moody v. NetChoice, LLC \- Wikipedia, https://en.wikipedia.org/wiki/Moody\_v.\_NetChoice,\_LLC

29. The NetChoice Cases: Will the Supreme Court Turn First, https://www.americanprogress.org/article/the-netchoice-cases-will-the-supreme-court-turn-first-amendment-law-on-its-head/

30. NetChoice Wins at Supreme Court Over Texas and Florida's, https://netchoice.org/netchoice-wins-at-supreme-court-over-texas-and-floridas-unconstitutional-speech-control-schemes/

31. Moody v. NetChoice, LLC (2024) | The First Amendment Encyclopedia, https://firstamendment.mtsu.edu/article/moody-v-netchoice-llc-2024/

32. In These Five Social Media Speech Cases, Supreme Court Set, https://www.eff.org/deeplinks/2024/08/through-line-suprme-courts-social-media-cases-same-first-amendment-rules-apply

33. A Small Step Toward Ending Jawboning, a Big Need for Disclosure, https://www.cato.org/blog/small-step-towards-ending-jawboning-big-need-disclosure

34. (PDF) Bugs in our Pockets: The Risks of Client-Side Scanning, https://www.researchgate.net/publication/355233857\_Bugs\_in\_our\_Pockets\_The\_Risks\_of\_Client-Side\_Scanning

35. \[2110.07450\] Bugs in our Pockets: The Risks of Client-Side Scanning, https://arxiv.org/abs/2110.07450

36. Encryption – EPIC – Electronic Privacy Information Center, https://epic.org/issues/cybersecurity/encryption/

37. Bugs in our pockets: the risks of client-side scanning, https://academic.oup.com/cybersecurity/article/10/1/tyad020/7590463

38. Fact Sheet: Client-Side Scanning \- Internet Society, https://www.internetsociety.org/resources/doc/2020/fact-sheet-client-side-scanning/

39. Carpenter v. United States \- Epic.org, https://epic.org/documents/carpenter-v-united-states-2/

40. Carpenter v. United States: Big Data is Different \-, https://www.gwlr.org/carpenter-v-united-states-big-data-is-different/

41. Facial Recognition and the Fourth Amendment in the Wake of, https://scholarship.law.uc.edu/cgi/viewcontent.cgi?article=1391\&context=uclr

42. ECtHR: Weakening encryption can violate human right to privacy, https://edri.org/our-work/european-court-of-human-rights-confirms-weakening-of-encryption-can-violate-the-human-right-to-privacy/

43. European Court of Human Rights Confirms: Weakening Encryption, https://www.eff.org/deeplinks/2024/03/european-court-human-rights-confirms-undermining-encryption-violates-fundamental

44. The ECtHR in Podchasov v. Russia – Preserving Encryption And, https://ohrh.law.ox.ac.uk/the-ecthr-in-podchasov-v-russia-preserving-encryption-and-denying-backdoors/

45. Podchasov v. Russia: the European Court of Human Rights, http://eulawanalysis.blogspot.com/2024/04/podchasov-v-russia-european-court-of.html

46. Podchasov v Russia: a new frontier in the crypto-wars before the, https://academic.oup.com/idpl/article/16/1/ipaf031/8371965

47. CASE OF PODCHASOV v. RUSSIA \- HUDOC, https://hudoc.echr.coe.int/eng?i=001-230854

48. As Russia Insulates Itself from Human Rights Bodies, State, https://advox.globalvoices.org/2015/12/18/as-russia-insulates-itself-from-human-rights-bodies-state-surveillance-decision-looms/

49. Zakharov v. Russia \- Wikipedia, https://en.wikipedia.org/wiki/Zakharov\_v.\_Russia

50. CASE OF ZAKHAROV v. RUSSIA \- HUDOC, https://hudoc.echr.coe.int/eng?i=001-77266

51. Encrypted Client Hello didn't solve censorship, but still may have a, https://adguard-dns.io/en/blog/encrypted-client-hello-misconceptions-future.html

52. QUICstep: Evaluating connection migration based QUIC censorship, https://arxiv.org/html/2304.01073v2

53. a universal way to bypass censorship using VPS, 3X-UI, Reality, https://habr.com/en/articles/990542/

54. Encrypted Client Hello (ECH) in Censorship Circumvention, https://corpus.lantern.io/papers/2025-niere-encrypted/

55. Resources for Digital Freedom in Iran | Javid Network Watch, https://javidnetworkwatch.com/resources

56. (PDF) Uncensored AI in the Wild: Tracking Publicly Available and, https://www.researchgate.net/publication/396718158\_Uncensored\_AI\_in\_the\_Wild\_Tracking\_Publicly\_Available\_and\_Locally\_Deployable\_LLMs

57. How to Run an LLM Locally on Your Mobile Phone with QVAC and, https://www.freecodecamp.org/news/how-to-run-an-llm-locally-on-your-mobile-phone-with-qvac-and-expo/

58. NTIA Supports Open Models to Promote AI Innovation, https://www.ntia.gov/press-release/2024/ntia-supports-open-models-promote-ai-innovation

59. NTIA advocates open AI model weights in report to White House, https://www.nextgov.com/artificial-intelligence/2024/07/ntia-advocates-open-ai-model-weights-report-white-house/398424/

60. The NTIA Report and California AI Bill: A New Era for Open Model, https://www.fenwick.com/insights/publications/the-ntia-report-and-california-ai-bill-a-new-era-for-open-model-governance

61. Compare Hintbox vs. SecureDrop in 2026 \- Slashdot, https://slashdot.org/software/comparison/Hintbox-vs-SecureDrop/

62. On unlinkability and denial of service attacks resilience of, https://www.researchgate.net/publication/348914620\_On\_unlinkability\_and\_denial\_of\_service\_attacks\_resilience\_of\_whistleblower\_platforms

63. Best GlobaLeaks Alternatives & Competitors \- SourceForge, https://sourceforge.net/software/product/GlobaLeaks/alternatives

64. Transparency & anti-corruption | MIAK — AI Shadow Government, https://miak.hu/en/teruletek/atlathato-korrupcioellenes/

65. Top SecureDrop Alternatives in 2026 \- Slashdot, https://slashdot.org/software/p/SecureDrop/alternatives

66. Nita Farahany QA on Cognitive Liberty Fall 2023, https://magazine.law.duke.edu/nita-farahany-qa-on-cognitive-liberty-fall-2023/

67. Technology and Security (TS) podcast | Neurotechnology, cognitive, https://www.ussc.edu.au/podcasts/technology-and-security-ts-podcast/neurotechnology-cognitive-liberty-and-information-warfare-with-professor-nita-farahany

68. Cognitive Liberty \- AI Ethics Lab – Rutgers University, https://aiethicslab.rutgers.edu/glossary/cognitive-liberty/

69. Your Right to Mental Privacy and the Impact in Cognitive Liberty, https://medium.com/@pedro.jutec/your-right-to-mental-privacy-and-the-impact-in-cognitive-liberty-ea5b97aada82

70. Mental Privacy in the Tech Age : talk on cognitive liberty, https://www.keynotespeak.com/keynote/talk-on-cognitive-liberty

71. Private Ownership Of Obscene Material Protected, https://www.annenbergclassroom.org/timeline\_event/private-ownership-obscene-material-protected/

72. Stanley v. Georgia | Case Brief, Ruling & Significance \- Study.com, https://study.com/academy/lesson/stanley-v-georgia-1969-case-brief-dissent-significance.html

73. arXiv:2409.05283v2 \[cs.CL\] 11 Oct 2024, https://arxiv.org/pdf/2409.05283

74. Alignment Tampering: How Reinforcement Learning from Human, https://arxiv.org/html/2605.27355

75. Political Bias Audits of LLMs Capture Sycophancy to the Inferred, https://arxiv.org/html/2604.27633v1

76. A Study on Explicit and Implicit Political Stereotypes in Large ... \- arXiv, https://arxiv.org/pdf/2510.08236

77. On the Relationship between Truth and Political Bias in Language, https://arxiv.org/html/2409.05283