# AI Services Censorship Report — April 2026 **Voidly Research — Report 2026-04** Published: 25 April 2026 License: CC BY 4.0 Permalink: https://voidly.ai/research/ai-censorship-april-2026 --- ## Lede Across **74 countries** with active OONI measurements between **12 April and 25 April 2026**, ChatGPT (`chatgpt.com`) is hard-blocked in 1 country and shows interference in 12 more. **Claude (`claude.ai`) is reachable from China** in the same measurement window — every probe in our China dataset that tested `claude.ai` returned a normal response, while every probe testing `chatgpt.com` failed. **Russia interferes with all five major Western AI services we tested** (chatgpt.com, chat.openai.com, openai.com, claude.ai, gemini.google.com), with anomaly rates between 3.8% and 8.1%. Gemini is the most-broadly-restricted AI service overall, showing block or interference signals in 17 countries. These asymmetries matter for the agent economy. An AI agent that depends on OpenAI's API will behave differently in China than one built on Anthropic's; a research tool that calls Gemini will fail in places where Claude works. The "internet is one thing" assumption baked into most LLM applications is not true, and the gaps are widening on a per-vendor basis. This report documents what we measured. Every claim links back to a live API call you can run yourself. --- ## Methodology Voidly ingests OONI's `web_connectivity` measurements every six hours via the OONI Aggregation API and records each domain × country × test outcome as an evidence row in our public database. For this report we selected five domains representing the dominant Western consumer-facing AI services and pulled every measurement event between 2026-04-12 and 2026-04-25. The full pull contains **3,604 measurement events** across **74 countries**, of which **233 events** in **23 countries** show either blocking (50) or interference (183). We classify each (country, service) cell using OONI's anomaly rate. A 100% anomaly rate on a sample of ten or more probes is reported as a hard block; anomaly rates above 50% on smaller samples are reported as partial blocks; anomaly rates below 50% are reported as interference (reachability is degraded but not absolute, often indicating ISP-specific blocking, blockpage delivery in some networks but not others, or transient enforcement). All other cells are reported as accessible. **Limitations.** OONI's coverage is uneven. Probe placement is not uniformly distributed across ISPs within a country, so a country-level "accessible" verdict can mask ISP-specific blocking, and a country-level "block" can be ISP-localized. We do not currently distinguish DNS poisoning from TCP-RESET from blockpage delivery. Provider-side geographic restrictions (where OpenAI or Anthropic terms-of-service exclude a country) are not captured by network-side measurement and so do not appear here as "blocks." Voidly's own probe network cross-checks a subset; we report only OONI-derived numbers in this report to keep the methodology auditable against a single upstream. --- ## Finding 1 — Country × service matrix 23 countries show some form of restriction on at least one of the five tracked AI services. | Country | ChatGPT | chat.openai.com | openai.com | Claude | Gemini | |---|---|---|---|---|---| | China (CN) | **BLOCK** (n=14) | **BLOCK** (n=12) | partial (n=1) | ok | **BLOCK** (n=14) | | Russia (RU) | INT 6.2% (n=14) | INT 8.1% (n=12) | INT 3.8% (n=11) | INT 4.1% (n=10) | INT 6.9% (n=14) | | Pakistan (PK) | INT 28% (n=5) | INT 5% (n=2) | ok | INT 4% (n=2) | partial (n=1) | | Canada (CA) | INT (n=4) | INT (n=8) | INT (n=4) | INT (n=5) | INT (n=8) | | Spain (ES) | INT (n=5) | INT (n=5) | INT (n=9) | INT (n=8) | INT (n=5) | | Germany (DE) | INT (n=1) | INT (n=3) | INT (n=3) | INT (n=2) | INT (n=4) | | United Kingdom (GB) | INT (n=2) | INT (n=1) | INT (n=1) | INT (n=1) | INT (n=1) | | United States (US) | INT (n=1) | INT (n=2) | INT (n=4) | INT (n=2) | INT (n=2) | | Kenya (KE) | INT (n=1) | ok | INT (n=1) | INT (n=1) | INT (n=1) | | Kazakhstan (KZ) | ok | INT (n=3) | ok | INT (n=1) | INT (n=1) | | South Africa (ZA) | partial (n=1) | ok | INT (n=1) | ok | partial (n=1) | | Brazil (BR) | INT (n=1) | ok | ok | INT (n=1) | INT (n=1) | | India (IN) | ok | ok | partial (n=1) | ok | partial (n=1) | | Nigeria (NG) | ok | ok | partial (n=1) | ok | ok | | Iraq (IQ) | ok | ok | ok | partial (n=1) | ok | | Algeria (DZ) | ok | ok | ok | partial (n=1) | ok | | Armenia (AM) | n/a | n/a | n/a | n/a | partial (n=1) | | Belarus (BY) | n/a | INT (n=1) | n/a | n/a | n/a | | Australia (AU) | INT (n=1) | n/a | n/a | n/a | n/a | | Japan (JP) | n/a | n/a | n/a | n/a | INT (n=2) | | Mexico (MX) | n/a | n/a | n/a | n/a | INT (n=1) | | Singapore (SG) | n/a | n/a | n/a | n/a | INT (n=1) | | Taiwan (TW) | n/a | n/a | n/a | n/a | INT (n=1) | --- ## Finding 2 — Most-restricted services Number of countries showing block/interference signals: ``` Gemini ████████████████████ 17 countries ChatGPT (chatgpt.com) ███████████████ 13 Claude (claude.ai) ███████████████ 13 openai.com (marketing) ██████████████ 12 chat.openai.com █████████████ 11 ``` Gemini's broader footprint reflects two effects: it shares Google's `*.google.com` certificate chain, and it is the only one of the five tracked services with hard blocks confirmed in two countries (China, plus partial blocks in Armenia, India, Mexico, Pakistan, Singapore, South Africa, Taiwan). --- ## Finding 3 — Vendor asymmetries - **China**: ChatGPT and Gemini are hard-blocked (100% anomaly rate on n=12–14 probes). Claude is reachable in the same measurement window with zero failures across 14 probes. - **Russia**: All five Western AI services show measurable interference (3.8–8.1% anomaly rate) — uniform across vendors, suggesting an ISP-level filter rather than domain-level block list. - **Pakistan**: ChatGPT shows the highest anomaly rate (28% on chatgpt.com, n=5). Claude and Gemini show much lower rates (4–5%). Filtering is OpenAI-asymmetric. --- ## Finding 4 — Blocking method breakdown Of 233 block/interference events, OONI's web_connectivity test reports an anomaly verdict but does not always disambiguate the underlying mechanism. Voidly's `blocking_method` column is empty for 100% of these AI-domain rows. Reconstructing methods (DNS poisoning vs TCP-RESET vs HTTP blockpage) requires going back to raw OONI measurement files. The China hard block is widely reported elsewhere as DNS poisoning combined with SNI filtering on the GFW; we have not independently verified that for this measurement window. Method-unattributed is a known limitation we flag rather than guess. --- ## Finding 5 — Measurement window All evidence: **2026-04-12 to 2026-04-25** (13-day window). We do not yet have a longitudinal AI-domain dataset of sufficient depth for "newly blocked this quarter." Quarterly trend coming in the July 2026 report. --- ## Finding 6 — AI access during election windows Two countries with restricted AI access in this window also have political tensions on Voidly's election-risk calendar: **Pakistan** (highest ChatGPT interference, 28%) and **Russia** (uniform interference across all services). We do not assert causation; the measurement window is short and we lack a pre-election baseline. --- ## Finding 7 — Coverage gap on emerging providers Perplexity, GitHub Copilot, Mistral, DeepSeek, Grok, Meta AI, Hugging Face, Cursor, Replicate, Cohere — OONI footprint is fewer than 10 events per domain across all countries in 13 days. Insufficient for accessibility claims at the same confidence as the big three. The censorship measurement community has historically focused on social, news, and messaging services; AI-specific monitoring is first-generation. --- ## Implications **For AI labs:** reachability is a product feature. China shows that vendor-by-vendor decisions get made independently, and a domain-naming choice (claude.ai vs chat.openai.com) can determine whether the service is reachable. **For AI-using businesses:** single-vendor AI dependencies are a per-country availability risk. Deployments should plan for graceful degradation when a vendor is unreachable in a given country. **For agent developers:** the agent economy assumes consistent reachability. It is not consistent. Agent frameworks should surface per-country availability of underlying model providers and treat AI-service unreachability as a recoverable error class with vendor failover. **For journalists and policy researchers:** everything is reproducible against OONI's public corpus. CC BY 4.0; cite it. --- ## Open data + reproduce **Live API:** ``` curl 'https://api.voidly.ai/v1/accessibility/check?domain=claude.ai&country=CN' curl 'https://api.voidly.ai/v1/accessibility/check?domain=chatgpt.com&country=CN' curl 'https://api.voidly.ai/v1/accessibility/check?domain=gemini.google.com&country=RU' ``` **Full dataset SQL (against Voidly evidence DB):** ```sql SELECT domain, country_code, signal_type, signal_value, COUNT(*) AS samples FROM evidence WHERE domain IN ('chatgpt.com','chat.openai.com','openai.com','claude.ai','gemini.google.com') AND signal_type IN ('block','interference') AND observed_at >= '2026-04-12T00:00:00Z' GROUP BY 1,2,3,4 ORDER BY 1,2; ``` JSON: https://voidly.ai/research/ai-censorship-2026-04/data.json --- ## About Voidly Voidly is a censorship measurement and intelligence platform. We aggregate public network-measurement data (OONI, CensoredPlanet, IODA) with our own global probe network into a queryable evidence database covering 168 countries and 5,300+ documented censorship incidents. Our data and APIs are free for researchers, journalists, and academics under CC BY 4.0. We also build privacy tools — an E2E-encrypted messenger, an open agent payment ledger, and a VPN — that operate inside the same network conditions our research describes. --- **Press contact:** press@voidly.ai **Web:** https://voidly.ai **Twitter/X:** @Voidly_ai **GitHub:** github.com/voidly-ai **API docs:** https://voidly.ai/api-docs **Cite as:** Voidly Research (2026). *AI Services Censorship Report — April 2026.* https://voidly.ai/research/ai-censorship-april-2026