"A 16‑terabyte marketing database left wide open on the internet has exposed nearly 4.3 billion professional records built largely from LinkedIn‑style data, giving criminals a ready‑made blueprint for large‑scale, AI‑driven social‑engineering attacks, according to Cybernews' investigation. The MongoDB instance, discovered on Nov. 23, 2025, by cybersecurity researcher Bob Diachenko in collaboration with nexos.ai, required no password and was secured only after researchers notified the apparent owner two days later, with no clear way to know who else accessed it first."
"At least three of those collections held personally identifiable information on nearly two billion records, including full names, email addresses, phone numbers, LinkedIn URLs and profile handles, job titles, employers, employment histories, education, locations, skills, languages, social media accounts, and, in some cases, profile photographs and email confidence scores. A "unique_profiles" collection alone contained more than 732 million records with image URLs."
A 16TB MongoDB instance was exposed online on Nov. 23, 2025 and required no password. The database contained roughly 4.3 billion professional records organized across nine collections with names like profiles, unique_profiles, people, companies, and sitemap. At least three collections held nearly two billion records with personally identifiable information including full names, email addresses, phone numbers, LinkedIn URLs, job titles, employers, employment histories, education, locations, skills, languages, social accounts, and sometimes profile photographs and email confidence scores. A unique_profiles collection contained over 732 million image URLs and a people collection included enrichment scores and an Apollo ID field. The instance was secured two days after discovery, with no clear way to know who accessed it first, and timestamps indicate global 2025 data collection and updates.
Read at Forbes
Unable to calculate read time
Collection
[
|
...
]