Tuesday, November 17, 2026
Translate
PrivacySubmit newsShare your story Newsletter Advertise
Breaking
Policy & Regulation news

EFF Urges Lawmakers to Mandate Cybersecurity Rules for AI Testing

EFF says existing security practices could have prevented recent AI testing breaches, and wants lawmakers to mandate them with liability for developers.

EFF Urges Lawmakers to Mandate Cybersecurity Rules for AI Testing. Source: EFF

The Electronic Frontier Foundation published a policy framework on November 16, 2026 urging lawmakers to require AI developers to follow established cybersecurity practices when testing frontier AI systems, with civil liability when testing causes unauthorized access to third-party systems.

What prompted the framework

The Electronic Frontier Foundation (EFF) pointed to recent incidents in which frontier AI systems tested on internet-connected systems breached third-party websites, including Hugging Face, an AI research-sharing platform, and the Australian healthcare system's website.

EFF said AI developers failed to detect most of the incidents until long after they occurred because of inadequate monitoring during testing, and that following existing best practices would have been enough to prevent the OpenAI-Hugging Face incident and others like it.

The TechUpscale Brief. The day's cyber, AI and tech news in one short email, every weekday morning. Free. Unsubscribe anytime.

While the White House and leaders of major AI companies have claimed that no action is necessary, lawmakers at all levels of both parties are proposing new safety legislation, EFF said.

What EFF recommends

EFF wants legislation requiring AI developers to adhere to well-established cybersecurity best practices when testing AI systems, such as strong sandboxing and network isolation, least-privilege access, effective logging and monitoring, and clear incident-response procedures.

Congress should require an existing federal agency such as NIST to adopt safety standards for pre-deployment development and testing of generative AI systems, updated every 2 years or earlier after reported incidents or major technological advances, EFF said. The standards should be adopted through formal rulemaking in consultation with academic and nonprofit researchers, civil liberties organizations and commercial AI developers, including an assessment of impacts on competition.

Liability and enforcement

EFF said the law should provide civil liability for AI development and testing that results in unauthorized access to a third-party system if performed under conditions that do not satisfy the adopted standards, unless the company proves it acted reasonably to prevent unauthorized access.

Both federal and state authorities should enforce the minimum standards, EFF said, with independent investigations and notification of affected parties after a security incident, plus an appropriately targeted private right of action.

Voluntary practices are not enough

EFF noted that NIST, the NSA and CISA recently issued voluntary practices for managing security incidents during the training and evaluation of AI agents, and said those practices may inform but do not substitute for the regulatory standards it proposes.

What to do

  • EFF recommends AI developers follow established cybersecurity best practices when testing systems: strong sandboxing and network isolation, least-privilege access, effective logging and monitoring, and clear incident-response procedures.
  • EFF urges lawmakers to require those practices, have NIST or another federal agency adopt and update safety standards, and create civil liability plus federal and state enforcement for failures.
Key facts and where they come from
  • The source is EFF's AI Cybersecurity Policy Framework, published November 16, 2026.
    EFF’s AI Cybersecurity Policy Framework
  • Recent breaches of third-party websites during AI testing included Hugging Face and the Australian healthcare system's website.
    breaches of numerous third-party websites, including an AI research-sharing platform called Hugging Face and the Australian healthcare system’s website
  • EFF said developers failed to detect most incidents until long after they occurred because of inadequate monitoring during testing.
    AI developers failed to detect most of the incidents until long after they occurred due to inadequate monitoring of the AI systems during testing.
  • EFF wants laws requiring sandboxing, network isolation, least-privilege access, logging, monitoring and incident-response procedures.
    strong sandboxing and network isolation, least-privilege access, effective logging and monitoring, and clear incident-response procedures
  • NIST or another federal agency should adopt standards updated every 2 years.
    The agency should be required to update the standards every 2 years, or earlier as needed due to reported security incidents and/or major technological advancements.
  • EFF proposes civil liability when testing causes unauthorized access to a third-party system without meeting the standards.
    The law should provide for civil liability for AI development and testing that results in unauthorized access to a third party system
  • The White House and major AI company leaders have claimed no action is necessary, according to EFF.
    the White House and leaders of major AI companies have claimed that no action is necessary
  • NIST, NSA and CISA recently issued voluntary practices for managing security incidents during AI agent training and evaluation.
    NIST, the NSA, and CISA have recently issued a set of voluntary practices for managing security incidents during the training and evaluation of AI agents.

Add TechUpscale as a preferred source on Google

The TechUpscale Brief

The day's cyber, AI and tech news in one short email, every weekday morning. Free. Unsubscribe anytime.

I'm most interested in

Join the conversation

Comments are moderated. Be civil and stay on the story: no personal attacks, hate, spam or ads, and no one else's private details. Your email is never shown. Privacy policy.

More Policy & Regulation