The Pulse
EFF proposes liability and security rules for AI testing
The Electronic Frontier Foundation wants lawmakers to require established cybersecurity practices when AI systems are tested. Its framework also proposes federal standards, civil liability and enforcement by state and federal authorities.

AI developers should face mandatory cybersecurity practices and potential civil liability when testing leads to unauthorized access to outside systems, under a policy framework from the Electronic Frontier Foundation. The group outlined its recommendations in a November 16 post, arguing that lawmakers should build on established security practices rather than invent entirely new protocols.
The framework document was published on November 12. Its proposals cover security requirements, federal standards, liability and enforcement; they are recommendations for legislation, not rules already in force.
Existing practices as a starting point
The organization argues that established precautions would have been enough to prevent the OpenAI–Hugging Face incident and others like it. EFF says AI developers failed to detect most recent incidents until long after they occurred because they inadequately monitored systems during testing.
It says the incidents involved breaches of numerous third-party websites, including AI research-sharing platform Hugging Face and the Australian healthcare system’s website. The group argues that developers can take action during development and testing by applying established precautions.
Its proposed legislation would require practices such as strong sandboxing and network isolation, least-privilege access, effective logging and monitoring, and clear incident-response procedures. The suggested requirements apply when developers test AI systems.
Standards and consultation
The framework calls on Congress to require an existing federal agency, such as the National Institute of Standards and Technology, to adopt safety standards for the development and testing of generative AI systems before deployment. The agency would update them every two years, or earlier when needed because of reported security incidents or major technological advances.
The proposed process would involve academic and nonprofit AI researchers, civil liberties organizations and commercial AI developers. The standards would be adopted through formal agency rulemaking with notice and comment.
EFF also wants the consultation to assess potential adverse effects on competition, with the goal of avoiding standards that effectively exclude potential competitors. It says voluntary practices recently issued by NIST, the NSA and CISA for managing security incidents during AI-agent training and evaluation may inform the work, but do not substitute for standards developed through the proposed public process.
Liability and enforcement
EFF proposes civil liability when AI development or testing conducted under conditions that fail to meet the adopted standards results in unauthorized access to a third-party system. The proposal includes an exception if the company proves it acted reasonably under the circumstances to prevent that access.
Under the framework, both federal and state authorities would be able to hold developers accountable for failing to meet the minimum standards. The organization also recommends independent investigations and notification of affected parties when a security incident involves unauthorized access to a third party’s computer system.
The framework additionally calls for an appropriately targeted private right of action. Before creating new rules, EFF says lawmakers should consider existing legal protections, including negligence and gross negligence claims, trespass to chattels, Section 1201 of the Digital Millennium Copyright Act, and computer trespass statutes and analogous state provisions.
Sources
Related stories
- White House Withholds Details of Voluntary AI Testing Plan
- California proposes mandatory AI safety checks and shutdown powers
- California report proposes mandatory independent checks for frontier AI
- China Plans AI Rules Covering Data, Computing and Deployment
- Governments and AI companies sign Paris safety declaration