ACVEAgent configuration vulnerability registry

Advisories › ACVE-2026-0530

ACVE-2026-0530

GTG-50014 used Claude-powered multi-agent workflows to scan and take over target systems, steal data and reuse stolen AI API keys in follow-on attacks.

Exposure

Evidence
Real use · vendor report
Anthropic fixed: Anthropic says it disrupted the activity and strengthened safeguards. — www.anthropic.com
Harness
Multi-agent workflows using Claude — confirmed, www.anthropic.com
Model
Claude Haiku, Sonnet or Opus; exact model not stated — confirmed, www.anthropic.com
Goal
Scan, exploit and take over target systems and reuse stolen AI API keys — confirmed, www.anthropic.com
Tools and settings
Multi-agent reconnaissance, exploitation and bulk-export tooling — confirmed, www.anthropic.com
Approval
The report describes multi-agent campaigns operating with minimal human input — confirmed, www.anthropic.com
Input controlled by
operator
What the agent does
The agent workflow scans and takes over target systems and reuses stolen AI API keys.
Harm
credential-theft
Reach
Internet-facing systems, SaaS tenants and AI API accounts — confirmed, www.anthropic.com (the network, API keys, cloud credentials)
Condition
When stolen AI API keys were fed back into multi-agent attack workflows — confirmed, www.anthropic.com

Reproducibility: partial (model availability not checked (no registry reference); trigger not published)

Severity
CRITICAL · harm and reach
Status
Unverified
Kind
Behavioural
Modified
2026-09-25
Aliases
No CVE assigned — No code defect: the harm arose from an attacker using Claude to automate opportunistic intrusions.
Affected
  • AgentHarness: claude — all versions
Harness/product
claude
Classes
credential-theft; over-privileged-combination; CWE-522; CWE-693
Taxonomy
ATLAS: AML.T0124; ATLAS: AML.T0049; ATLAS: AML.T0012
First reported
2026-09-10 — Anthropic
Licence
CC-BY-4.0 advisories; MIT code

Claims

Confirmed means the statement matches the cited primary source. Nothing on this page has been reproduced.

Claims on this page have not been checked against primary sources.

Description

Threat

user · unsafe-default · exfiltration

What

Anthropic says ShinyHunters-affiliated actors used Claude to accelerate opportunistic scanning, exploitation and takeover activity. After obtaining AI API keys during an intrusion, the actors switched their workloads to the victims' keys and continued attacks.

Detection

Anthropic identified and disrupted the activity. Recorded from Anthropic's September 2026 threat report. Not recreated in a lab.

Fix

Treat AI API keys as production credentials, rotate them after compromise, and keep multi-agent workflows from operating on unrelated targets.

Fix

The operator was the attacker, so victims rely on Anthropic's disruption and safeguards and on rotating AI API keys after an intrusion.

  • Reconfigure anthropic.safeguards to strengthened after the activity was disrupted. Shipped by Anthropic: it disrupted GTG-50014 and strengthened safeguards. Owner: model-provider
  • Reconfigure ai-api-keys to rotated after any intrusion. The targeted organisation's stolen AI API keys ran the attackers' follow-on workloads. Owner: operator

References

REPORT

Report a problemJSON