Home › Guides › Tor Exit Relays After the October 2026 Incident
✓ checked 11 October 2026 · sources: Tor Metrics and Onionoo (data CC0), Tor Project forum posts · method · caveats · download the data
The number of Tor relays carrying the Exit flag fell from 3,013 on 4 October 2026 to 2,715 on 5 October, a one-day drop of 298 (9.9%). On 5 October the Tor Project told relay operators that it was flagging affected relays and that the directory authorities were rejecting them. As of 2026-10-11, 2,703 running relays hold the Exit flag.
Read this first. Tor Metrics shows the drop, but it doesn’t say why. We link the timing to the incident because the Tor Project and an affected operator group describe the same events; we haven’t verified who was affected or how many relays were involved. Tor exit relays are run by volunteers and organisations for legitimate privacy use. This page describes the size and spread of the network, not any individual operator.
exit relays running, 11 Oct 2026 2,703one-day drop, 4 to 5 Oct 9.9%exit selection probability held by the largest country (NL) 29.3%
Headline numbers
- Exit relays fell 353 (11.5%) between 1 and 5 October, from 3,068 to 2,715. On the same day, 4 to 5 October, all running relays fell by 234. By 9 October the count was 2,702.
- The exit count was already well below its peak. Tor Metrics shows 4,704 relays with the Exit flag on 4 Feb 2026, so the 9 October figure is 42.6% lower. Exits were 1,008 on 1 January 2020 and 2,947 on 1 January 2026.
- The network is concentrated in a few countries and organisations. By Onionoo’s exit selection probability, the Netherlands carries 29.3%, Germany 24.8% and the US 16.3%. The US has the most exit relays (1,017, 37.6% of all) but many are small.
- Three autonomous systems carry 50.6% of exit selection probability. The largest, registered as “Church of Cyberology”, has 476 exit relays and 21.7%. An autonomous system is a network, not necessarily one operator.
- Relay families: operators can declare relays as a family so that Tor never uses two of them in one circuit. The largest family of exits holds 149 relays and 10.3% of exit selection probability, among 499 families or single relays. The ten single relays with the highest probability hold 3.0% between them.
- 230 exit relays (8.5%) run a Tor version that Onionoo does not mark as recommended, and 1,338 also carry the Guard flag.
Exit relays per day, 25 September to 9 October 2026
Where exit capacity sits
| Country | Exit relays | Exit selection probability |
|---|---|---|
| NL | 605 | 29.3% |
| DE | 332 | 24.8% |
| US | 1,017 | 16.3% |
| SE | 211 | 9.5% |
| AT | 123 | 6.1% |
| FR | 65 | 3.2% |
| NO | 45 | 2.1% |
| DK | 15 | 1.8% |
Largest autonomous systems
| AS | Registered name | Exit relays | Exit selection probability |
|---|---|---|---|
| AS215125 | Church of Cyberology | 476 | 21.7% |
| AS60729 | Stiftung Erneuerbare Freiheit | 260 | 18.1% |
| AS62744 | Quintex Alliance Consulting | 600 | 10.8% |
| AS208323 | Foundation for Applied Privacy | 120 | 6.0% |
| AS198093 | Foreningen for digitala fri- och rattigheter | 148 | 5.6% |
| AS205100 | F3 Netze e.V. | 16 | 3.5% |
| AS61125 | SABOTAGE LLC | 31 | 3.4% |
| AS41281 | KeFF Networks Ltd | 37 | 3.3% |
What the Tor Project and operators say
In a post to the tor-relays list, a Tor Project staff member wrote that “Over the weekend we received reports that relays of some of our operators were intercepting SSH connections.” The post adds: “For now, we have been flagging those affected relays and the directory authorities are rejecting them from the network.” An operator group that runs relays under the name Quetzalcoatl later published its own timeline, saying two of its servers were rooted with the Ebury OpenSSH backdoor and that an open exit policy “had been live for about 52 hours on two Guard+Exit relays moving roughly 3.4 TiB a day”. These are the participants’ accounts; we haven’t checked them.
For how this connects to proxies and scraping traffic, see Where Bot Traffic Comes From, our Anti-Bot Census and the residential proxy botnet timeline.
Russian users and Tor bridges
Updated 15 October 2026: added this section on bridge users. The exit-relay figures above are from 11 October and haven’t changed.
Bridges are unlisted Tor relays that help people connect where Tor is blocked. Tor Metrics estimates that 95,811 of the 226,128 people connecting through bridges on 12 October 2026 were in Russia (42.4%). Iran was next with 31,529 and the US third with 27,069. The same day, Tor Metrics estimates 21,498 users in Russia connecting directly to public relays, so about 4.5 times as many Russian users came in through bridges.
- Up 18.0% on a year earlier. Russian bridge users averaged 92,678 a day from 1 to 12 October 2026 (12 days with data), against 78,564 in October 2025 and 53,409 in August 2025.
- The highest daily figure since January 2025 was 106,627, on 27 April 2026.
- Russia’s share of bridge users is lower than at the start of 2025 (50.7% in January 2025), because bridge use elsewhere grew too.
- Onionoo listed 2,070 running bridges on 2026-10-15.
Tor Metrics doesn’t say why use changes. Press reports describe pressure on VPNs in Russia: The Riposte, a newsletter, reported on 13 October 2026 that a VPN project’s channel had reported problems with its keys on 4 and 21 September, which the newsletter called “a new wave of attempts by Roskomnadzor” to block VPNs, including those using the VLESS protocol. We haven’t verified those reports. For a separate compatibility problem in tools that use VLESS with REALITY, see the post-quantum REALITY split.
Tor Metrics user numbers are estimates based on directory requests, and some days have no estimate (624 of the 650 days from 1 January 2025 to 12 October 2026 have a Russian bridge figure). Monthly averages use the days with data.
Method
Data. (1) Tor Metrics relayflags.csv, daily counts of relays by flag, 1 January 2020 to 9 October 2026, downloaded 11 October 2026. (2) For the bridge section, Tor Metrics userstats-bridge-country.csv and userstats-relay-country.csv, downloaded 15 October 2026, and Onionoo’s running-bridge summary. (3) Onionoo /details?running=true&flag=Exit, published 2026-10-11 16:00:00 UTC and fetched 11 October 2026: 2,703 running exit relays with country, autonomous system, family, version and exit probability.
Calculations. Shares are sums of Onionoo’s exit_probability, which it describes as only “an approximation” of path selection. Families join any relays that list each other in effective_family. Country is the country Onionoo assigns to the relay.
Verification. A separate script recomputes every number with its own code (a different grouping method for families) and checks each quote against the saved Tor Project pages.
Caveats
- A fall in the Exit flag count has several possible causes, including relays being removed by directory authorities, operators shutting relays down and routine churn. Tor Metrics doesn’t attribute the change.
- Exit selection probability is an estimate, and country and AS data come from geolocation and registry information that can be wrong.
- The exit-relay figures come from Onionoo’s public relay list, which doesn’t include bridges.
- Counts of relays aren’t counts of users, and a high share of exit probability doesn’t mean a single party can see traffic.
Download the data
- Daily exit-flag and running-relay counts (CSV), 2020 to 9 October 2026.
- Exit relays by country and autonomous system (CSV), Onionoo snapshot of 2026-10-11.
- Daily bridge and relay users for Russia (CSV), Tor Metrics estimates, 1 January 2025 to 12 October 2026.
Cite as: ProxyPicker, “Tor Exit Relays After the October 2026 Incident”, 11 October 2026, https://proxypicker.com/tor-exit-network/. Data: Tor Project, Tor Metrics and Onionoo (CC0).
Related: Where Bot Traffic Comes From · Anti-Bot Census · Residential proxy botnet timeline · Post-quantum REALITY split.
Press & data requests: email [email protected] for interviews, the raw data behind this page, or corrections. See also our contact page.