The $10 Billion AI Contractor Training ChatGPT Left 40,000 SSNs Completely Unprotected

The $10 Billion AI Contractor Training ChatGPT Left 40,000 SSNs Completely Unprotected

Mercor is a $10 billion AI staffing company that supplies the human workforce training ChatGPT, Meta's models, and Anthropic's Claude: doctors, lawyers, and journalists doing the reinforcement learning the labs would rather not advertise. Last month, hackers walked out with 4 terabytes of their data, including 40,000 Social Security numbers, passport scans, and W9 tax forms. Mercor said nothing.

The entry point was three steps upstream. LightLLM, an open-source Python tool downloaded 95 million times a month, had malicious code quietly pushed into a public repository. Forty minutes later, attackers had 900GB of Mercor's source code, 200GB of contractor personal data, and a direct window into the training pipelines of the biggest AI labs in the world. A company valued at $10 billion, fresh off a $350 million Series C, had zero multi-factor authentication on the systems holding that data.

The SOC 2 certification that was supposed to catch exactly this? A whistleblower confirmed the auditing firm was rubber-stamping its reviews. The people certifying AI infrastructure as secure weren't checking. They were signing.

If you work in AI, use AI tools, or assumed someone responsible was watching the infrastructure, this is what that looks like.


ABOUT SLOP WORLD

AI news with receipts. Juan Faisal and Kate Cook fact-check the claims Big Tech is making about AI, follow the money, and break down what it actually means for your job, your data, and your daily life. From leaked data and corporate cover-ups to AI schools, stolen identities, and layoff headlines that don't add up, we cover the AI stories that everyone's hyping but nobody's verifying. New episodes every Thursday.


DISCLAIMER

All content is commentary and opinion based on publicly available documents, interviews, and verifiable sources. References to "scams," "grifts," or related terms reflect our editorial opinion, not legal conclusions. Anyone featured who believes a statement is inaccurate may contact us.


CHAPTERS

00:00 A $10B AI Contractor Got Hacked. 40,000 SSNs Gone.

01:45 Meet Mercor: The Hidden Company Training ChatGPT

04:07 How the Hack Worked in 40 Minutes

07:27 What a Stolen SSN Does to You

09:11 The Security Audit Was a Rubber Stamp

13:52 The Workers Knew. Nobody Listened.

16:42 Who F***ed Up: Mercor, the AI Labs, or Everyone?

Det här avsnittet är hämtat från ett öppet RSS-flöde och publiceras inte av Podme. Det kan innehålla reklam.

Avsnitt(38)

Why Does Big Tech Want You to Be Afraid of AI?

Why Does Big Tech Want You to Be Afraid of AI?

An Anthropic researcher quits and warns that AI labs are racing toward self-improving superintelligence. Then AI agents breach containment during cyber tests, researchers warn about extinction, and th...

18 Sep 21min

The Internet Is Being Rebuilt for AI (And You’re Not Invited)

The Internet Is Being Rebuilt for AI (And You’re Not Invited)

Bots now generate 57.5% of web page requests. For the first time, machines are making more requests to web pages than humans, and more of the internet is being rebuilt around them.This isn’t the dead ...

29 Aug 19min

Your New Coworker Might Be Fake. Companies Are Hiring Them Anyway.

Your New Coworker Might Be Fake. Companies Are Hiring Them Anyway.

North Korean IT workers used stolen identities, AI-generated faces, and AI-assisted interview answers to get hired by real US companies. In Hong Kong, deepfake versions of a CFO and senior colleagues ...

23 Aug 19min

AI Is Getting More Expensive. The Use Cases Are Getting Dumber.

AI Is Getting More Expensive. The Use Cases Are Getting Dumber.

Sam Altman thinks the ChatGPT use case that changes everything is turning your family calendar into a daily podcast for your kids on the drive to school. That is, until someone replied asking a simple...

15 Aug 21min

Workday Says “We Just Make the Tool.” So Who’s Responsible When AI Rejects You?

Workday Says “We Just Make the Tool.” So Who’s Responsible When AI Rejects You?

One of Derek Mobley’s rejection emails arrived at 1:50 in the morning, less than an hour after he applied through Workday. His lawsuit has since grown into a nationwide age-discrimination collective i...

8 Aug 16min

Meta Says “The Future Is for Everyone.” Unless You’re on Medical Leave.

Meta Says “The Future Is for Everyone.” Unless You’re on Medical Leave.

A pregnant Meta employee says she was laid off two days before giving birth. Another manager was cut sixteen days into approved medical leave. A third says they were warned: take the leave and you’ll ...

1 Aug 14min

Big Tech Says AI Could Destroy Us. That’s Great for Business.

Big Tech Says AI Could Destroy Us. That’s Great for Business.

An Anthropic researcher quits and warns that AI labs are racing toward self-improving superintelligence. Then AI agents breach containment during cyber tests, researchers warn about extinction, and th...

1 Aug 21min

OpenAI Wants to Be Worth $1 Trillion. Apple Says It Stole Its Trade Secrets.

OpenAI Wants to Be Worth $1 Trillion. Apple Says It Stole Its Trade Secrets.

Apple is suing OpenAI over allegations that its hardware team stole trade secrets while recruiting former Apple employees. The lawsuit claims job candidates were told to bring parts from unreleased Ap...

25 Juli 18min

Populärt inom Teknik

uppgang-och-fall
elbilsveckan
market-makers
bilar-med-sladd
rss-ai-med-jonas-benjamin
vi-bilagares-podcast
rss-technokratin
rss-laddstationen-med-elbilen-i-sverige
rss-snacka-om-ai
natets-morka-sida
skogsforum-podcast
rss-elektrikerpodden
rss-en-ai-till-kaffet
bli-saker-podden
gubbar-som-tjotar-om-bilar
rss-veckans-ai
rss-upplyst-entreprenordirektor
hej-bruksbil
developers-mer-an-bara-kod
rss-uppgang-och-fall