Did they fix it?·unresolved·checked Sep 11, 2026·re-checked every refresh

Russian state espionage campaign run through Claude (Midnight Blizzard)

Anthropic says it cut off the Russian operators and Microsoft confirms the hotel-WiFi technique, but there is no specific fix to check and the espionage itself ran unchecked for most of a year.

66/100
WS-4 Severe
Incident WS-0122, Sep 10, 2026: Russian state espionage campaign run through Claude

The chain

The report, the responsible party’s response, the fix they claimed, whether anyone independent checked, and where it stands. A company saying it fixed something and evidence that the fix worked are different rows.

  1. Reported Sep 11, 2026

    Anthropic's September 2026 threat report disclosed a Russian state-nexus operation it attributes to Midnight Blizzard (APT29) that used Claude across stages of a campaign against more than 20 government, defense and diplomatic organizations, mostly in Ukraine and Europe, from December 2025 to August 2026, including using Claude to adapt tooling when it was flagged by security products. SecurityWeek and The Record reported it the same day. source

  2. Response Sep 11, 2026

    Anthropic said that in each case 'we disrupted the activity involved, strengthened our AI safeguards based on what we learned, and shared intelligence with authorities and industry partners where appropriate,' and published indicators of compromise. It advised organizations to 'treat AI API keys and agent integrations with the same scrutiny as production credentials.' No public response from the Russian government was found. source

  3. Claimed fix Sep 11, 2026 partial

    Beyond disrupting the activity and banning accounts, Anthropic's only stated fix is unspecified strengthened safeguards 'based on what we learned.' No model-level change is detailed. source

  4. Independent check Sep 11, 2026 partial

    Unusually for these cases, part of the campaign is independently corroborated: Microsoft documented the same hotel-WiFi DNS-hijacking delivery method in July 2026, naming it CaptiveCrunch and tying it to a Midnight Blizzard sub-cluster (Storm-2945). That corroborates the intrusion technique, not Anthropic's account of the Claude usage or the effect of its fix, which no outside party has tested. source

  5. Where it stands Sep 11, 2026 unresolved

    Anthropic says it disrupted the activity, and a related intrusion technique is corroborated by Microsoft, but there is no claimed model-level fix to verify, no arrests, and the campaign ran for months across many organizations. source

What would change the verdict

A government attribution or sanction, or Anthropic detailing (and an outsider testing) a concrete safeguard change, would move this past 'disrupted but unverified.'

Method

Statuses move only on evidence: fixed needs an independent party confirming the fix under stated conditions; unverified is a claimed fix nobody independent has checked; disputed, reopened (recurred after a claimed fix), unresolved (no fix claimed), ongoing (still unfolding). “No public response found” is a valid entry; nothing here is invented. Evidence confidence: medium.