AI Security

    'Sleeper Agent' Tactic Hides Payloads in Open-Weight AI Models

    A tiny share of a neural network may conceal malware, raising questions about detection.

    Malware can lurk inside just 0.00001% of a neural network. How, then, can any user tell it's there?

    DR

    Dark Reading

    Source

    Dark Reading
    'Sleeper Agent' Tactic Hides Payloads in Open-Weight AI Models
    AI Analysis

    Generated insight by OnAirToday AI

    If malicious payloads can occupy such a tiny fraction of a model, ordinary inspection may struggle to reveal them. This raises questions about how users can assess the safety of open-weight models before using them.

    Powered by AI • Not a substitute for human analysisMore Insights →
    Read the full story at Dark Reading →
    Share
    Filed under:AI Security

    Comments

    Keep comments relevant and respectful. No links, advertising, self-promotion, or spam.

    0/1000