AI startup that Mark Zuckerberg paid $14 billion 'leaked' sensitive data of Google, Elon Musk’s xAI and Facebook: What the company has to say

Scale AI, the

AI startup

that is a key partner for companies like

Meta

and xAI, is facing scrutiny after a report revealed that the company has been exposing sensitive client and contractor data through publicly accessible

Google Docs

. The findings, released by Business Insider, raise serious questions about

Scale AI

's security practices and its commitment to

client confidentiality

.The report uncovered thousands of pages of project documents across 85 individual Google Docs, some containing highly confidential information related to Scale AI's work with major tech clients.

Documents show information on training data related to Google and Elon Musk’s xAI

These documents included details on how Google utilised ChatGPT to refine its Bard chatbot, as well as at least seven “confidential” instruction manuals from Google outlining issues with Bard and how contractors should address them.For Elon Musk's xAI, public documents revealed specifics of "Project Xylophone," including training materials and 700 conversation prompts aimed at enhancing the AI's conversational abilities on diverse topics.

Similarly, Meta training documents, marked confidential, were found publicly accessible, containing links to audio files demonstrating "good" and "bad" speech prompts for their AI products.Contractors reported that despite attempts at codenaming, it was often easy to identify the client they were working for due to the nature of the tasks or even explicit company logos embedded in presentations. Some even said they could determine the client simply by prompting the AI model directly.

Contractor data also exposed: Report

Beyond client information, Scale AI also left sensitive personal data belonging to thousands of its contractors exposed in unsecured Google Docs. Spreadsheets, readily viewable by anyone with the URL, listed private Gmail addresses, work performance details, and even categorized some workers as “high quality” or suspected of “cheating.”

Scale AI responds to report on 'leaking' data

In response to the publication’s findings, Scale AI stated it takes

data security

seriously and is investigating the matter.“We are conducting a thorough investigation and have disabled any user's ability to publicly share documents from Scale-managed systems,” a spokesperson said. "We remain committed to robust technical and policy safeguards to protect confidential information and are always working to strengthen our practices," the spokesperson added.The revelations come on the heels of Meta's blockbuster investment in Scale AI. However, clients including Google, OpenAI, and xAI reportedly paused work with Scale AI following Meta's investment.