SOCRadarAmeer Owda
BabidiGhost Used AI to Target European Banks and Insurers
Part of the CVE-2023-38646 story · since 20 Nov
- Fri 20 Nov, 09:30 UTC. SOCRadar, first report. BabidiGhost Used AI to Target European Banks and Insurers. This article, 1st of 1.
SOCRadar Research has documented BabidiGhost, a French-speaking operator who used a jailbroken AI agent to target European banks, insurers, payment providers and municipalities between 6 and 14 November 2026. The operation began with an exposed staging server in France where the agent’s configuration and extensive prompt logs were recovered, allowing SOCRadar to reconstruct the end-to-end campaign without engaging any victims’ systems.
The operator's instructions were in plain, misspelled French, while the agent generated the offensive tooling and exfiltration scripts, relaying thousands of model prompts during nine days of activity. The operation culminated when the model API key was revoked on 14 November 2026, after which the agent went silent, though implants persisted for about another day.
BabidiGhost is the second operator observed using the BlackHatSect0r AI toolchain; the first, BlackHatSect0r, was the subject of Part I in September 2026. The BabidiGhost setup relied on the same agent and jailbreak framework but showed markedly lower technical skill.
Across more than 50 targets in Western Europe, the campaign yielded four primary outcomes for 14 organisations under scrutiny: the insurer suffered customer-d database exfiltration via an internet-facing Metabase instance vulnerable to CVE-2023-38646; a municipal storage server exposed to the internet with a weak admin password led to the theft of 23 archives (about 8.6 GB)
from 13 municipalities; six banks were profiled with varying degrees of exposure, from validated accounts to blocked access behind IP allowlists; and several payments, crypto and betting sites yielded mostly aborted attempts or non-viable results. The report highlights that ordinary controls—MFA, IP allowlisting and server-side validation—stopped the AI agent in several cases, while the toolkit enabled rapid discovery of exposed data, readable secrets and unauthenticated endpoints.
How the story developed
1 article from 1 outlet since 20 Nov. Still developing; the latest report was 11 hours ago. This was the first report.
- Fri 20 Nov, 09:30 UTC. SOCRadar, first report. BabidiGhost Used AI to Target European Banks and Insurers. This article, 1st of 1.
- CVE not in KEV
- No CVE
- Larger: an outlet’s first report
- Coverage over time
Show all 1 articles as a listHide the list
- Fri 20 NovDay 01
- BabidiGhost Used AI to Target European Banks and Insurers09:30 · SOCRadar · First report · This article, 1st of 1