当前位置:首页 > 焦點

【】

"All [Rabbit] R1 responses ever given can be downloaded," according to an R1 research group called Rabbitude.

Rabbit and its R1 AI device has already been dunked on for being nothing more than an Android app wrapped up in a hardware gadget, but something much more alarming is afoot.

SEE ALSO:I tested Rabbit R1 vs. Meta AI: The winning AI assistant will surprise you

The report (via The Verge) said Rabbitude gained access to the codebase and discovered API keys were hardwired into its code. That means anyone with these keys could "read every response every r1 has ever given, including ones containing personal information, brick all r1s, alter the responses of all r1s [and] replace every r1’s voice." The investigation discovered that these API keys are what provided access to ElevenLabs and Azure for text-to-speech generation, Yelp for reviews, and Google Maps for location data.

Mashable Light SpeedWant more out-of-this world tech, space and science stories?Sign up for Mashable's weekly Light Speed newsletter.By signing up you agree to our Terms of Use and Privacy Policy.Thanks for signing up!

What's worse, Rabbitude said it identified the security flaw on May 16 and that Rabbit was aware of the issue. But "the API keys continue to be valid as of writing," on June 25. Continued access to the API keys means bad actors could potentially access sensitive data, crash the entire rabbitOS system, and add custom text.

The following day (June 26) Rabbit issued a statement on its Discord server saying that the four API keys Rabbitude identified have been revoked. "As of right now, we are not aware of any customer data being leaked or any compromise to our systems," said the company.

But the plot thickens. Rabbitude also found a fifth API key that was hardwired in the code, but not publicly disclosed in its investigation. This one is called sendgrid, which provides access to all emails to the r1.rabbit.tech subdomain. At the time Rabbitude published its follow-up report, the sendgrid API key was still active. Access to this API key meant Rabbitude could access additional user information within the R1's spreadsheet functions and even send emails from rabbit.tech email addresses.

If you were already skeptical of the R1's half-baked capabilities that Mashable Tech Editor Kimberly Gedeon blamed on "rushed innovation, disillusionment, and impetuousness" in her review, this might be your sign that Rabbit is at best, not worth the money, and at worst, incapable of keeping your data private.

TopicsArtificial IntelligencePrivacy

分享到: