The artificial intelligence start -up that recently won an investment of $ 14 billion in Meta has an incredibly Janky “safety system – using public Google Doc files to store confidential information on customers like Meta, Google and XAI, according to a report.
It was reported earlier this month Meta agreed to take a 49% stake in the AI scale for $ 14.8 billion And bring the CEO of the startup, Alexandr Wang, to lead a new “superintelligence” laboratory.
This scandalously high price indicates that Meta thinks that Wang and her business are essential to pass the IA division of the social media society to the next level.
But the company was strangely relaxed with regard to its work with high-level customers, leaving top secret projects and sensitive information such as email addresses and payment of details in Google Docs accessible to anyone with a link, According to Business Insider.
“We are carrying out an in-depth survey and have disabled any user’s ability to publicly share systems documents managed by Scale,” an AI spokesperson for The Post said.
“We remain attached to solid technical and political guarantees to protect confidential information and we are still working to strengthen our practices.”
Although there is no indication that public files have led to a violation, they could leave the company sensitive to hacks, according to cybersecurity experts.
Google and Xai did not immediately respond to requests for comments from the position. Meta refused to comment.
Five AI entrepreneurs on a current and old scale told BI that the use of Google Docs was widespread throughout the company.
“The entire Google Docs system always seemed incredibly Janky,” said a worker.
Bi said he had been able to see thousands of pages of project documents on 85 Google Docs detailing the sensitive scale of the AI scale with Big Tech customers, as well as the way Google used the Openai Chatppt to refine its own chatbot.
At least seven Google manuals marked “confidential”, including recommendations to improve the chatbot, then calculated Bard, were accessible to the public, according to the report.
Google Doc public files included details on Elon Musk’s “Xylophone Project”, such as training documents with 700 conversation prompts to improve a chatbot’s conversation skills, according to the report.
The so-called “confidential” meta-training documents with audio clips of “good” and “bad” speeches to train its AI products were also left public.
Although these secret projects have often received code names, several AI entrepreneurs on a scale said it was always easy to determine for which customer they worked.
Certain documents related to code name projects evenly included the company logo, such as a presentation which included a Google logo, according to BI.
When you work with AI products, the chatbot sometimes revealed the customer when asked, the entrepreneurs said.
There were also Google Doc calculation sheets accessible to the public who listed the names and email addresses private of thousands of workers, according to the media.
A spreadsheet was frankly entitled “Good and bad” and labeled dozens of workers like “high quality” or “cheating”, said Bi. Another document reported workers with “suspicious behavior”.
According to the report, more public documents have more public documents have established the amount of individual entrepreneurs and included detailed tickets on disputes and remuneration differences, according to the report.
