Close Menu

    Stay Ahead with Exclusive Updates!

    Enter your email below and be the first to know what’s happening in the ever-evolving world of technology!

    What's Hot

    China Just Blocked Meta From Buying One of Its Most Promising AI Startups. The Tech Cold War Just Got More Concrete.

    May 10, 2026

    A Security Flaw in Hugging Face’s Robotics Platform Shows What Happens When Open-Source AI Begins Controlling Physical Machines

    May 10, 2026

    AI Is Making Identity the Most Dangerous Attack Surface in Cybersecurity. CrowdStrike’s $740 Million Bet Proves It.

    May 10, 2026
    Facebook X (Twitter) Instagram
    Facebook X (Twitter)
    PhronewsPhronews
    • Home
    • Big Tech & Startups

      China Just Blocked Meta From Buying One of Its Most Promising AI Startups. The Tech Cold War Just Got More Concrete.

      May 10, 2026

      Google Just Entered Nvidia’s Business by Selling Its Own AI Chips to Customers for the First Time. The Competitive Map Has Changed.

      May 8, 2026

      The EU Just Forced Smartphone Makers to Bring Back Replaceable Batteries — Here’s What That Means for Your Next Phone

      May 7, 2026

      Everyone Is Racing to Pure Fusion But This Startup Thinks That’s the Wrong Bet. Here’s Why It’s Mixing Fusion and Fission Instead

      May 7, 2026

      AWS Is Growing at 28% But Amazon Is Losing the AI Cloud War to Microsoft and Google

      May 5, 2026
    • Crypto

      Market Collapse: What Happened to NFTs?

      April 23, 2026

      Quantum Computing Advances Force Coinbase and Institutional Custodians to Rethink Crypto Security

      March 8, 2026

      AI Assisted Hacking Groups Target Crypto Firms With Multi-Layered Social Engineering

      February 18, 2026

      Global Crypto Regulations Expand as 2026 Begins With New Data Collection Frameworks and National Laws

      January 16, 2026

      Coinbase Bets on Stablecoin and On-Chain Growth as Key Market Drivers in 2026 Strategy

      January 10, 2026
    • Gadgets & Smart Tech
      Featured

      Foldable Phones Are No Longer a Gimmick — The Motorola Razr 2026 Is the Latest Sign That Foldables Are Going Mainstream

      By fariehanMay 3, 2026
      Recent

      Foldable Phones Are No Longer a Gimmick — The Motorola Razr 2026 Is the Latest Sign That Foldables Are Going Mainstream

      May 3, 2026

      Meta Raises Quest VR Headset Prices as Component Costs Rise

      May 1, 2026

      Robotics Showcase: China Uses a Half-Marathon to Signal Progress in Humanoid Tech

      April 27, 2026
    • Cybersecurity & Online Safety

      A Security Flaw in Hugging Face’s Robotics Platform Shows What Happens When Open-Source AI Begins Controlling Physical Machines

      May 10, 2026

      AI Is Making Identity the Most Dangerous Attack Surface in Cybersecurity. CrowdStrike’s $740 Million Bet Proves It.

      May 10, 2026

      Businesses Are Already Running AI Agents They Cannot Secure. Palo Alto Acquiring Portkey Is the Industry Admitting It.

      May 8, 2026

      An AI Agent Deleted an Entire Company’s Database in Seconds. This Is No Longer a Safety Debate. It Is Now a Business Liability Crisis

      May 5, 2026

      Project Glasswing: How Anthropic Is Trying to Keep Its Most Dangerous Model in Check

      April 26, 2026
    PhronewsPhronews
    Home»Artificial Intelligence & The Future»A Security Flaw in Hugging Face’s Robotics Platform Shows What Happens When Open-Source AI Begins Controlling Physical Machines
    Artificial Intelligence & The Future

    A Security Flaw in Hugging Face’s Robotics Platform Shows What Happens When Open-Source AI Begins Controlling Physical Machines

    preciousBy preciousMay 10, 2026No Comments
    Facebook Twitter Pinterest LinkedIn WhatsApp Reddit Tumblr Email
    Share
    Facebook Twitter LinkedIn Pinterest Email
    Photo Credit: The Hacker News

    Robotics labs operate on a basic assumption that the machines inside them take instructions only from authorized sources. However, that assumption immediately falls apart when the software running those machines can be reached and hijacked by anyone with a network connection and the right payload. That is exactly the situation researchers have now exposed with LeRobot, Hugging Face’s open-source robotics platform.

    A critical vulnerability, tracked as CVE-2026-25874 with a CVSS score of 9.3, was publicly disclosed on April 28, 2026. The flaw allows unauthenticated attackers to execute arbitrary code on systems running LeRobot, with no login credentials required. The platform has nearly 24,000 stars on GitHub.

    What the Flaw Does

    The vulnerability sits in LeRobot’s asynchronous inference pipeline, where the platform uses Python’s pickle.loads() function to deserialize data received over gRPC channels. Those channels have no Transport Layer Security and require no authentication, meaning any attacker who can reach the server over a network can send a crafted payload and trigger code execution on the host machine.

    The attack path runs through specific gRPC handlers, including SendPolicyInstructions and SendObservations, which process raw byte streams and pass them through pickle before any validation runs. Because deserialization happens first, a malicious payload executes before the system checks whether the incoming object is even the right type.

    And when a successful exploit happens, it gives an attacker full control of the PolicyServer host. From there, they can steal API keys, SSH credentials, and proprietary model files, move laterally across internal networks, corrupt machine learning models, and in production environments where LeRobot controls physical robots, manipulate or disrupt those machines directly.

    The Irony Baked Into the Code

    It is also important to note that Hugging Face previously developed Safetensors, a serialization format built specifically to eliminate the security risks that come with using pickle. LeRobot does not use it. Instead, the codebase uses pickle.loads() and includes nosec comments placed directly next to those calls. Those comments instruct automated security linting tools to stay quiet, even when the tools correctly flag a problem.

    Security researcher Valentin Lobstein, who publicly disclosed additional information about the flaw, described the contradiction plainly, saying, “Hugging Face created Safetensors – a serialization format designed specifically because pickle is dangerous for ML data. And yet their own robotics framework deserializes attacker-controlled network input with pickle.loads(), with # nosec comments to silence the tool that was trying to warn them.”

    How Long the Team Has Known

    A private report about the same flaw was submitted in December 2025 by a researcher using the alias “chenpinji.” The LeRobot team responded in early January 2026, acknowledging that parts of the codebase needed significant refactoring. Steven Palma, the project’s tech lead, confirmed that deployment security had not been a priority during the platform’s research phase. A fix is planned for version 0.6.0, but no release date has been confirmed.

    Until a patch is available, security experts recommend restricting network access to LeRobot instances and using firewalls or VPNs to limit exposure to trusted networks only.

    The LeRobot case is a clear signal that as open-source AI frameworks move from research labs into environments where they influence or control physical systems, building in security from the start is not optional.

    Artificial Intelligence CVE-2026-25874 exploitation GitHub Open Source Security Trends Hugging Face Hugging Face LeRobot vulnerability Open-source robotics security 2026 Unauthenticated Remote Code Execution (RCE)
    Share. Facebook Twitter Pinterest LinkedIn Tumblr Telegram Email
    precious
    • LinkedIn

    I’m Precious Amusat, Phronews’ Content Writer. I conduct in-depth research and write on the latest developments in the tech industry, including trends in big tech, startups, cybersecurity, artificial intelligence and their global impacts. When I’m off the clock, you’ll find me cheering on women’s footy, curled up with a romance novel, or binge-watching crime thrillers.

    Related Posts

    China Just Blocked Meta From Buying One of Its Most Promising AI Startups. The Tech Cold War Just Got More Concrete.

    May 10, 2026

    AI Is Making Identity the Most Dangerous Attack Surface in Cybersecurity. CrowdStrike’s $740 Million Bet Proves It.

    May 10, 2026

    Businesses Are Already Running AI Agents They Cannot Secure. Palo Alto Acquiring Portkey Is the Industry Admitting It.

    May 8, 2026

    Comments are closed.

    Top Posts

    Coinbase responds to hack: customer impact and official statement

    May 22, 2025

    Anthropic Will Use Claude User Chats For Data Training

    October 16, 2025

    MIT Study Reveals ChatGPT Impairs Brain Activity & Thinking

    June 29, 2025

    Cursor AI Hits 1 Million Daily Users. Why Developers Are Switching to This Coding Tool

    March 23, 2026
    Don't Miss
    Artificial Intelligence & The Future

    China Just Blocked Meta From Buying One of Its Most Promising AI Startups. The Tech Cold War Just Got More Concrete.

    By preciousMay 10, 2026

    China’s top economic planning agency has ordered Meta to walk away from its acquisition of…

    A Security Flaw in Hugging Face’s Robotics Platform Shows What Happens When Open-Source AI Begins Controlling Physical Machines

    May 10, 2026

    AI Is Making Identity the Most Dangerous Attack Surface in Cybersecurity. CrowdStrike’s $740 Million Bet Proves It.

    May 10, 2026

    Businesses Are Already Running AI Agents They Cannot Secure. Palo Alto Acquiring Portkey Is the Industry Admitting It.

    May 8, 2026
    Stay In Touch
    • Facebook
    • Twitter
    About Us
    About Us

    Evolving from Phronesis News, Phronews brings deep insight and smart analysis to the world of technology. Stay informed, stay ahead, and navigate tech with wisdom.
    We're accepting new partnerships right now.

    Email Us: info@phronews.com

    Facebook X (Twitter) Pinterest YouTube
    Our Picks
    Most Popular

    Coinbase responds to hack: customer impact and official statement

    May 22, 2025

    Anthropic Will Use Claude User Chats For Data Training

    October 16, 2025

    MIT Study Reveals ChatGPT Impairs Brain Activity & Thinking

    June 29, 2025
    © 2025. Phronews.
    • Home
    • About Us
    • Get In Touch
    • Privacy Policy
    • Terms and Conditions

    Type above and press Enter to search. Press Esc to cancel.