Opinion Piece: AI Changes Technology, Not the Humility of Criminal Law

Recently, a video featuring AI-synthesized voices of political figures has sparked controversy, with public discussion expanding from the content itself to whether police visits for verification infringe upon freedom of expression. According to public reports, after receiving a complaint, authorities investigated a video suspected of using AI deepfake technology to mimic the president’s voice. The purpose was to clarify the facts and production process, without conducting searches, seizures, or any coercive measures restricting personal freedom. However, creators argue that police involvement in creative activities creates a chilling effect akin to 'checking the water meter.'

This incident transcends the individual case, touching upon the boundaries of state power. Governments may investigate AI content suspected of illegality, and individuals cannot claim complete immunity from legal scrutiny simply by citing technological creativity. Nevertheless, initiating criminal procedures must still be based on concrete suspicion of criminal activity and actual harm to protected legal interests. If criminal procedures intervene prematurely due to political sensitivity or public pressure—before legal evaluation is complete—the principle of criminal law as a 'last resort' risks eroding. The scope of penal authority may then expand from responding to crime to managing social disputes and political conflicts, posing new challenges to the limits法治国 places on public power.

In this case, the key issue is not whether AI was used, but whether the conduct reaches a level warranting criminal intervention. Merely using AI to synthesize voices is not inherently illegal—technology is simply a tool for creation and information dissemination. What the law must assess is the manner of use and its effects: whether the content could mislead the public into believing it is an authentic statement, and whether it harms public credibility, reputation, or other legally protected interests.

If the content constitutes political commentary, parody, satire, or secondary creation, and is clearly recognizable as such by the audience, it should generally be protected under freedom of speech. Democratic societies allow sharp, exaggerated, and even offensive political expression. Criticism and satire directed at those in power have traditionally enjoyed heightened constitutional protection. Offensiveness alone is insufficient grounds for criminal intervention.

Conversely, if AI-generated content convincingly mimics official statements by political leaders or government agencies—such as fabricating the voices of the president, premier, police, or disaster response units—and spreads information that significantly influences public judgment, the harm extends beyond individual reputation. It undermines public information order, governmental credibility, electoral fairness, and societal trust.

Thus, legal intervention is justified not because political voices are imitated, but because someone’s identity is exploited to create misleading information that causes tangible legal harm. Legal evaluation must focus on the act and its consequences, not on the political status or power of the subject portrayed.

Police verification following a complaint is a standard step in criminal procedures. However, whether such action complies with the principle of proportionality depends not only on the absence of formal coercive measures like search or seizure, but also on the real-world impact of state intervention. A visit by criminal investigators carries strong symbolism of state authority. To most people, it is difficult to immediately distinguish between voluntary cooperation, being questioned, and formal criminal investigation. When criminal procedures enter a creator’s personal space before any legal determination of guilt, many may reconsider whether it is still worthwhile to publish controversial content.

The chilling effect is not primarily about actual punishment, but about the quiet narrowing of public discourse. When creators begin to think certain criticisms should not be voiced, certain political topics avoided, or certain works left unpublished—even without any legal judgment—self-censorship emerges. A法治国 must not only refrain from arbitrarily punishing citizens but also avoid imposing procedural pressures that lead people to retreat before any law is broken. Especially regarding political speech, procedures must not be perceived as an extension of content censorship.

Of course, freedom of speech is not an absolute shield against legal accountability. If all AI-generated content were exempt from liability, deepfake technology would become easier to exploit for fraud, disinformation, and impersonation of public authority. Democratic societies protect free expression while also safeguarding public trust. Content that criticizes policies—even with sharp, sarcastic language—should be constitutionally protected. But using deepfake technology to impersonate official entities, create misleading information, and damage public trust must be subject to legal regulation.

What requires restraint is the exercise of state penal power. People’s speech is often uncomfortable, filled with exaggeration, satire, and criticism. The maturity of a democratic system lies not in universal agreement with all speech, but in the state’s willingness to uphold法治国 boundaries and refrain from hastily deploying criminal power, even in the face of controversial expression.

AI technology changes how information is generated, but it does not alter the threshold for criminal law intervention. Regardless of technological advancement, penal authority must be grounded in concrete suspicion, harm to legal interests, and fulfillment of criminal elements—not on political sensitivity or public opinion trends.

*The author is an adjunct lecturer in law and cybersecurity at a higher education institution.