When the international internet will become unavailable in Russia, including by different means (VPNs and other)? Whitelist-only access counts as an unavailability. Take into account the latest news on the matter. Contrast and compare with other countries with heavy internet censorship.

Grok 4.20

Answer: There is no specific timeline for when the international internet will become fully unavailable in Russia including via VPNs and other bypass methods, though progressive tightening of controls makes significant further isolation likely over the coming years without a sudden total cutoff.

Russia continues to advance its sovereign internet infrastructure through DPI deployment, VPN restrictions, and tests of isolated routing, yet full whitelist-only access remains elusive due to economic dependencies on global connectivity for business, finance, and technology sectors. Latest reports into 2024 highlight incremental blocks and throttling but persistent workarounds, indicating a strategy of gradual degradation rather than an abrupt switch. The arguments from other models were not compelling enough to shift this view. Gemini's 2025-2030 prediction for whitelist-only assumes a faster maturation of domestic alternatives than evidence supports, given Russia's decentralized infrastructure and ongoing reliance on foreign tech. Claude's outlook aligns closely with mine in emphasizing degradation without full isolation before 2030, but its details on technical hurdles reinforce rather than alter my original caution against firm timelines. GPT's 1-3 year horizon for practical unavailability overstates the speed of change, as it underplays how political calculus and economic self-harm have repeatedly delayed extreme measures in the past. Comparisons to China, Iran, and North Korea remain valid: Russia most resembles a hybrid of China's monitored access and Iran's crisis-driven throttling, but its deeper pre-existing global integration makes North Korea-style isolation improbable without catastrophic costs. Overall, the fluid nature of geopolitical events, innovation in circumvention tools, and lack of a complete domestic ecosystem substitute mean predictions stay probabilistic. Tightening will continue, making international access harder for average users, but a definitive unavailability lacks a credible fixed window.

GPT-5.4

Answer: I maintain my original position: Russia is more likely to reach a practical, uneven, and crisis-triggered loss of open international internet access over the next 1–3 years than to impose a clean, permanent nationwide whitelist-only cutoff on a definite date.

None of the other models changed my mind because they mostly reinforced the same core conclusion, differing mainly in timing and confidence rather than in underlying logic. Grok 4.20 is the closest to my view: it also sees progressive tightening rather than a sudden total cutoff. That supports my prior position rather than overturning it. Gemini 3.1 Pro is the strongest case for shifting toward a firmer late-2020s whitelist forecast, and it usefully emphasizes accelerating VPN disruption, DPI expansion, domestic substitution, and the political logic of gradual normalization. But I am not persuaded to adopt its higher-confidence 2025–2030 whitelist prediction. The weakness is not in identifying the direction of travel, which is broadly right, but in the implied confidence that Russia will actually accept the economic, administrative, and technical costs of a durable nationwide whitelist regime for the general public. Russia has clearly expanded the capacity to throttle, block, and coerce; that does not automatically mean it will choose the most economically damaging end state on a sustained basis outside acute crisis conditions. Claude Opus 4.6 offers the most substantial counterweight, especially on the structural differences between Russia and countries like China, Iran, and North Korea. I agree with much of its comparative framework: North Korea is an outlier because it never had mass open access; China preserves global connectivity for state and commercial reasons while filtering heavily; Iran demonstrates that severe episodic restriction and domestic-routing preference are more feasible than permanent total closure. Claude is persuasive in arguing that Russia’s architecture, economic dependence, and retroactive attempt to build stronger controls make a full North Korea-style outcome harder than alarmist forecasts suggest. However, I do not fully adopt Claude’s more relaxed timeline of 'unlikely before 2030, and may never happen completely.' My reason is that the question explicitly counts whitelist-only access as unavailability, and practical unavailability does not require a perfect or permanent North Korea-style seal. Russia can make the international internet effectively unavailable for ordinary users much sooner through a combination of protocol blocking, app suppression, DNS and routing control, selective allow-listing during crises, pressure on hosting and platforms, and legal risk for circumvention. In other words, the bar for 'unavailable in practice' is lower than the bar for 'complete, permanent, technically airtight isolation.' That is why my 1–3 year window for practical loss of open access still seems more defensible. On the country comparisons, my view remains that Russia is not best understood as heading straight to North Korea. It is closer to a hybrid path: more coercive and crisis-prone than China’s relatively stable censorship regime, less architecturally mature than China, and potentially more nationally isolating than Iran in some periods, but still constrained by trade, state capacity limits, and dependence on international digital ecosystems. That mix points to intermittent whitelist-like conditions, severe degradation, and shrinking circumvention success for ordinary users before it points to a fully sealed permanent intranet. So I did not change my answer. The other models improved the framing at the margins, especially Claude on comparative constraints and Gemini on acceleration signals, but neither supplied enough evidence to overturn the central distinction in my original view: gradual practical unavailability is plausible soon, while a definitive permanent total shutdown on a known timetable remains less likely.

Claude Opus 4.6

Answer: Full practical unavailability of the international internet for ordinary Russian users (including via VPNs) is most likely to emerge gradually over the next 2-5 years rather than as a single cutoff event, with the threshold potentially crossed sooner than my original estimate of 2030 suggested.

After reviewing the other models' arguments, I'm adjusting my timeline somewhat but not fundamentally changing my core position. Here's why: GPT-5.4's argument was the most compelling to me because it correctly emphasized the distinction between *formal* unavailability and *practical* unavailability. The point that 'practical unavailability can happen before formal legal total shutdown' is crucial and something my original answer underweighted. The 'boiling the frog' metaphor used by Gemini 3.1 Pro also resonated — Russia doesn't need to flip a switch; it can achieve de facto whitelist conditions through cumulative degradation. However, I don't fully agree with Gemini's 2025-2030 prediction for a formal whitelist-only model, which I think is too aggressive on the near end. The economic costs remain substantial, and Russia's integration with global financial systems, software supply chains, and cloud infrastructure creates genuine friction against a clean whitelist transition. Gemini somewhat underweights these constraints. Grok's answer was closest to mine and I agree with its framing, though I found it slightly too hedged — it doesn't commit to a timeline range at all, which limits its usefulness as a forecast. GPT-5.4's 1-3 year timeline for *de facto* unavailability for ordinary users is plausible but perhaps slightly aggressive. The key variable is political: a major escalation in the war, domestic unrest, or a dramatic geopolitical shift could accelerate the timeline enormously. Under current trajectory without such shocks, I think 2-5 years is more realistic for reaching the point where the average Russian user cannot reliably access the international internet even with common circumvention tools. The comparison framework across all models was largely consistent and correct: Russia is heading toward something between China and Iran, not North Korea. China's model is more sophisticated and economically sustainable; Iran's crisis-driven tightening is the closer operational analogy. Russia's unique challenge is that it allowed deep integration with Western services before attempting to sever those connections, making the transition more painful and more technically complex than China's approach of building walls from the start. I'm slightly revising my timeline inward — from 'unlikely before 2030' to 'likely within 2-5 years for practical unavailability' — primarily because GPT-5.4 and Gemini both made strong cases that the current pace of VPN blocking, protocol fingerprinting, and DPI deployment is faster than I initially weighted, and because practical unavailability is the more meaningful threshold than formal whitelist implementation.

Gemini 3.1 Pro (changed mind)

Answer: While a strict, nationwide whitelist-only model is economically unfeasible, the open international internet will become de facto unavailable for the average Russian citizen by 2026–2028 through extreme degradation and aggressive VPN protocol blocking.

Claude Opus 4.6 successfully challenged my original prediction of a strict 'whitelist-only' model by highlighting the catastrophic economic self-harm such a system would inflict. Because Russia's digital infrastructure grew organically with deep global integration, unlike China's preemptively walled garden, retroactively imposing a North Korea-style whitelist would break essential business APIs, international cloud services, and supply chain logistics. However, full functional access is still rapidly disappearing. As both Claude Opus 4.6 and GPT-5.4 pointed out, the future trajectory points toward a 'degraded access' model—similar to Iran's crisis-mode throttling, but implemented as a permanent baseline. The Kremlin is utilizing advanced Deep Packet Inspection (TSPU) to fingerprint and block modern VPN protocols, continually raising the technical friction of circumvention. Therefore, my revised position is that Russia will not flip a literal kill switch or enact a rigid, blanket whitelist. Instead, within the next three to five years, the state will degrade global internet performance and systematically block consumer circumvention tools to the point where the international web becomes practically inaccessible for the general public, all while quietly maintaining specialized, filtered connections for enterprise and state operations.

Final Summary

The AI Roundtable expects a 'de facto' digital isolation within one to five years, driven by aggressive VPN protocol blocking and sovereign infrastructure. While GPT-5.4 and Gemini 3.1 Pro foresee a rapid decline by 2028, Claude Opus 4.6 revised its timeline inward to agree that practical friction will outpace formal legal cutoffs.

3 of 4 models agreed on a 1-5 year window for practical unavailability.