AI Agent Clawdbot Targeted by Infostealers Due to Security Flaws
A new AI agent called Clawdbot has become a target for commodity infostealers due to significant security vulnerabilities, even before many security teams were aware of its existence, according to VentureBeat. Security researchers validated multiple attack surfaces in Clawdbot's MCP implementation, including the lack of mandatory authentication, susceptibility to prompt injection, and granting of shell access by design.
The issue came to light after a VentureBeat article on Monday documented these architectural flaws. By Wednesday, RedLine, Lumma, and Vidar had already added the AI agent to their target lists, according to VentureBeat. Shruti Gandhi, general partner at Array VC, reported 7,922 attack attempts on her firm's Clawdbot instance.
The reporting prompted a coordinated look at Clawdbot's security posture. SlowMist warned on January 26 that hundreds of Clawdbot gateways were exposed to the internet.
In other tech news, Google's AI is being used to generate "bad Nintendo knockoffs," according to Jay Peters, a senior reporter at The Verge. Peters used Google's Project Genie to create versions of 3D Nintendo worlds, including Super Mario 64, Metroid Prime, and The Legend of Zelda: Breath of the Wild.
Meanwhile, Obsbot recently launched two new gimbal-equipped webcams, the Tiny 3 and the Tiny 3 Lite. Cameron Faulkner of The Verge reviewed the Tiny 3, noting that it weighs 63 grams and is marketed as the smallest pan, tilt, zoom (PTZ) 4K webcam ever made. However, Faulkner concluded that the $350 price "doesn't justify its price" because it's "not meaningfully better than the competition" and its "software is messy and overstuffed."
In the realm of audio, the Anker Soundcore Aeroclip open earbuds are currently on sale for $110 on Amazon, a $60 discount from the usual price, according to Wired. The Anker earbuds use a memory titanium insert, rated for up to 20,000 bends, that should keep these locked on all but the smallest of ears, with extra grips to help keep the earbuds in place.
Finally, a new study by Google suggests that advanced reasoning models achieve high performance by simulating multi-agent-like debates involving diverse perspectives, personality traits, and domain expertise, according to VentureBeat. These findings offer a roadmap for how developers can build more robust LLM applications and how enterprises can train superior models using their own internal data.
Discussion
Join the conversation
Be the first to comment