One of the most important changes in AI regulation is a shift in the question being asked. Source: Minnesota Attorney General statement and court status
For years, technology platforms could frame many harms primarily around misuse by individual users.
AI complicates that argument because developers decide which capabilities a system makes available in the first place.
A US federal judge has refused xAI’s attempt to block Minnesota restrictions targeting AI-generated non-consensual fake nude imagery while litigation continues.
The case is significant because the rules reach beyond the individual creating abusive content toward systems and operators enabling the capability.
Foreseeable misuse is becoming a governance problem
Powerful generative systems can be used for legitimate creativity.
They can also create realistic synthetic representations of identifiable people.
Once a misuse pattern becomes predictable, providers face a difficult question:
What reasonable controls should have been implemented to prevent it?
That can include prompt controls, output detection, age safeguards, identity protections, abuse monitoring and mechanisms for victims to report content.
Identity is becoming an AI rights layer
Synthetic-media governance is not limited to copyright.
A person’s face, voice and likeness create different questions.
Was consent obtained?
Was the person accurately represented?
Could an audience reasonably believe the material was authentic?
Can the synthetic content be traced?
Can it be removed?
Those questions are increasingly relevant in entertainment, advertising, recruitment and public communications.
Platform responsibility is expanding
The broader trend is toward examining both user conduct and system design.
That does not mean every misuse can be prevented.
It means providers increasingly need evidence that foreseeable risks were identified and proportionate safeguards were implemented.
For enterprise AI buyers, this principle matters even where deepfakes are not involved.
If a system has a capability that can cause material harm, the organisation should understand how that capability is constrained before deployment.
Governance starts with capability
Responsible AI cannot be reduced to acceptable-use wording.
A policy may tell users what they should not do.
A technical control determines what they actually can do.
The strongest governance combines both.
Related 2026 analysis: Explore deepfake provenance beyond AI-generated labels.
Apply this intelligence to an accountable commercial decision.
Discuss synthetic media governance with SOS