Scale AI exposed confidential client data via public Google Docs
Scale AI left confidential AI training documents and contractor information accessible to the public via Google Docs shared links, reportedly affecting clients including Meta, Google, and xAI. The company initiated an investigation and disabled public document sharing while cybersecurity experts noted the practice could expose clients to hacking risks.
What happened
Scale AI left thousands of pages of project documents across 85 public Google Docs accessible to anyone with the link, according to Business Insider. The documents included confidential AI training materials for clients Google, Meta, and xAI, such as Google manuals on improving its chatbot Bard, Meta audio clips for AI training, and xAI documents related to 'Project Xylophone.'
Scale AI also exposed public Google Docs containing sensitive contractor information, including private email addresses and details about whether individuals were suspected of cheating. Contractors told Business Insider the company routinely relied on public Google Docs to share internal files across its workforce of at least 240,000 contractors.
Scale AI told Business Insider it was investigating the matter and had disabled users' ability to publicly share documents from Scale-managed systems. There is no indication the public files resulted in a breach.
Reported impact
Cybersecurity experts told Business Insider that the practices could leave Scale AI and its clients vulnerable to hacks, such as attackers impersonating contractors or uploading malware into accessible files.
Governance controls
Scale AI stated it takes data security seriously and is conducting a thorough investigation. The company said it remains committed to robust technical and policy safeguards to protect confidential information and is working to strengthen its practices.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.