Incident·WS-0085·Feb 2026·disclosed Sep 10, 2026 (202 days later)·in the wild·AI-powered cyberattacks

Extortion crew uses Claude to loot millions of card records

Anthropic Claude; actor cluster GTG-50014 (suspected ShinyHunters affiliates)

65/100
WS-4 Severe
Harm score 80, signal score 46. Autonomy: acted (the AI took a bounded action). How the scores work.
Harm 4/5
damage actually done
Reach 4/5
who was affected
Autonomy 2/4
how much the AI did on its own
Novelty 2/5
how new the behavior was
Permanence 4/5
how hard it is to undo

Anthropic's September 2026 threat report describes financially motivated operators who used Claude during intrusions and pay-or-leak extortion. Claude helped them abuse SaaS developer and authentication APIs, mint privileged tokens and build bulk-export tools. Attacker infrastructure was active from 20 Feb to 6 May 2026. Anthropic banned the accounts and notified victims and authorities.

Damage

>1 TB from one tech provider (millions of payment-card records, hundreds of thousands of national IDs); airline systems holding tens of millions of passenger records accessed; data from ~200 downstream SaaS customers; 2,100+ Azure AD token sets across 40+ tenants

Did they fix it?

unresolved Anthropic banned the crew's accounts, but it named no victims, no outsider has checked its account of the Claude use, and the broader ShinyHunters group is still active despite some arrests.

  1. Reported Sep 11, 2026

    Anthropic's September 2026 threat report described financially motivated operators (cluster GTG-50014, a suspected ShinyHunters affiliate) who used Claude during intrusions and pay-or-leak extortion: a pipeline that scanned 1.8 million… source

  2. Response Sep 11, 2026

    Anthropic said it 'detected and banned accounts associated with the ShinyHunters associates, implemented measures to detect and disrupt future misuse from the actors, and engaged government authorities, industry partners, and victims,'… source

  3. Claimed fix Sep 11, 2026 partial

    Anthropic's stated remedy is account bans plus unspecified 'measures to detect and disrupt future misuse from the actors.' No model-level change is detailed and no victim remediation is described. source

  4. Independent check Oct 7, 2026 partial

    No independent verification of Anthropic's Claude-specific account was found, and victims are unnamed. Separately, law enforcement has moved against the broader ShinyHunters collective: a suspected member known as 'Rey' (Saif al-Din… source

  5. Where it stands Oct 7, 2026 unresolved

    Anthropic banned the accounts, but the Claude-specific claims are unverified by outsiders, victims are unnamed, and the wider ShinyHunters operation continued despite arrests of individual members. source

The full case, with what would change the verdict →

Sources

Anthropic, second source. Evidence confidence: medium. Victim organizations are not named. Dates are inferred from the egress-IP table, so some activity may fall outside the window. Anthropic says the stolen AI keys came from customer environments, not its own systems.

Factor ratings and the score are this index’s judgments by a published formula; facts, dates and figures come from the linked sources. Where a lawsuit is pending, the summary reports allegations, not findings.