A look at GFWeb, which tracks millions of domains to uncover the Great Firewall's domain blocklists; GFWeb data reveals China blocked ChatGPT on March 2, 2023
Rest of World received exclusive access to a platform that tracks patterns and timing of Chinese online censorship.
Context & Ripple Effects
GFWeb adds domain-level observability to a censorship system whose reach had previously been documented at scale: university researchers found the Great Firewall blocked roughly 311,000 domains in 2020, including many apparently accidental blocks (earlier measurement of large-scale domain blocking).
The platform also places AI-service access within a longer pattern of network controls, from the Gmail blocking episode to apparent efforts against TLS-based circumvention tools. Later coverage of substantially different access denial in Henan suggests the effects of those controls need not be uniform across China (evidence of regional variation in website denial).
First-order effects
- GFWeb gives researchers and the public a way to inspect the timing and patterns of domain blocks rather than infer them only from individual access failures.
- The data identifies ChatGPT as unavailable through the Great Firewall from March 2, 2023, limiting direct access for users connecting from China.
Second-order effects
- A dated, domain-level record makes it easier to distinguish broad filtering patterns from isolated outages or accidental blocking, raising the evidentiary bar for assessing censorship changes.
- Developers and organizations that depend on externally hosted AI services must treat network reachability as a separate constraint from a service’s technical availability.
Third-order effects
- If independent measurement platforms continue to mature, Great Firewall policy can become more contestable and comparable over time, including across regions and categories of online services.
- The case points toward AI access being governed not only by model providers and product rules, but also by national network gatekeepers—a form of government-gated AI access whose practical impact can vary by location.
The trend: AI services are increasingly subject to national network-level access controls, making independent measurement of digital gatekeeping more consequential.