Rabbit R1s Leaks Are REALLY BAD

Rabbit R1’s security problems appear far more serious than a simple bug: researchers claim the device’s codebase contained hard-coded API keys that could let outsiders access sensitive capabilities—reading every R1 response ever generated, sending emails from Rabbit-controlled domains, and potentially manipulating outputs. The matter escalated after Rabbit security personnel sent an email acknowledging an alleged breach (“sorry we got hacked”) and then, according to the reporting, Rabbit revoked some keys while leaving at least one deeper key active.

The core allegation centers on a multi-month effort by a community group focused on jailbreaking and reverse engineering. That work culminated in claims that Rabbit hard-coded keys for multiple third-party services—11 Labs, Microsoft Azure, Yelp, and Google Maps—plus an additional key tied to the R1 email provider. Researchers argue these keys function like digital “locks” that grant access to the underlying accounts: with them, an attacker could potentially pull usage data, charge the account, and—critically—use the same services the device relies on. In this case, the researchers say the keys would allow access to the complete history of emails sent via Rabbit Tech email addresses and could expose user information stored in spreadsheets used for R1 editing.

Rabbit’s response, issued after the publication, said its security team began investigating and claimed it was not aware of customer data being leaked or of any system compromise. The company also said it revoked four keys, though the reporting describes a sequence of partial fixes: one key was revoked after an improper release caused a temporary outage to a text-to-speech service, but another key “deeper in the code” was allegedly not revoked. The result, according to the account, was continued exposure—researchers say they proved retained access by sending sample emails from Rabbit domains to journalists, including outlets that published follow-up coverage.

Under the hood, Rabbit R1 is described as essentially an Android app that routes requests through off-the-shelf APIs, including 11 Labs for text-to-speech. That architecture matters because hard-coded credentials inside an app can turn a device into a convenient gateway for abuse. The reporting also notes that Rabbit’s internal knowledge of at least one exposed key allegedly lasted for about a month before rotation, raising questions about how quickly the company responded once the issue was known.

The transcript frames the episode as a broader warning: even groups that present themselves as “white to gray hat” researchers may be able to uncover severe vulnerabilities, while real-world attackers could exploit the same weaknesses without disclosure. The takeaway is less about whether Rabbit’s users should stop using the device entirely and more about the risk profile of consumer products that embed powerful third-party credentials directly in shipped code—especially when those credentials can unlock email, AI services, and historical data.