AI-specific criminal-code amendments

Movement organisations pursue narrow, harm-specific criminal-law amendments creating new criminal offences for named AI-enabled acts: non-consensual synthetic intimate imagery (deepfake sex-abuse imagery), AI-generated child sexual abuse material, AI-generated election-interference impersonation of candidates, AI-facilitated identity theft, AI-generated harassment and threats. The vehicle is a specific bill in a national or sub-national legislature (US DEFIANCE Act, US TAKE IT DOWN Act, UK Online Safety Act deepfake provisions, EU AI Act criminal-adjacent provisions, various state-level deepfake statutes) that adds a specific criminal offence, defines its elements narrowly around the AI-enabled specifics, sets penalties, and designates enforcement authority. The strategy converts a diffuse AI harm into a specific prosecutable act.

An actor chooses this strategy because criminal law creates individual deterrence on the perpetrator side of the harm, provides survivors a specific vindicating remedy (a prosecution with a defendant on trial), gives law-enforcement agencies a specific tool their existing statutes do not adequately cover, and produces a public-record judgment (a conviction, a sentence) that names the wrong and rules on it. On specific AI harms — non-consensual intimate imagery, AI-generated CSAM — criminal legislation has also been dramatically easier to pass than any other AI regulation, because the survivor-led coalitions building the case are joined by law-enforcement lobbies, victims-of-crime advocacy, and cross-partisan alignment that comprehensive AI regulation has not commanded. The strategy also produces early legislative wins the movement can point to when arguing for the broader regulatory frameworks it wants.

It trades systemic reach for offence-specific narrowness. Criminal-code amendments punish individual perpetrators and typically leave platform liability, vendor liability, and upstream tool-provider liability untouched — the deepfake tool remains legal, the platform hosting the material may be Section-230-immune from the criminal offence attaching to the poster, and the model provider whose product enabled the harm faces no direct criminal exposure. The strategy is also entangled with civil-liberties concerns from adjacent movements: criminal-law expansion is contested by prison-abolition, prosecutorial-discretion, and free-speech coalitions on principled grounds, and coalition tension is a permanent feature of any AI-criminal-code campaign the movement runs. Enforcement variance is also structural: a criminal statute that exists but is not prosecuted, or is prosecuted selectively by jurisdiction, produces a formal legal win with limited deterrent effect on the ground.

Ecology

Distinct from coalition lobbying of binding regional regulation. Regional-regulation lobbying pursues comprehensive administrative-law frameworks — the EU AI Act's risk tiers, comprehensive US federal AI legislation — that establish civil obligations on operators and enforcement powers with regulatory agencies. This strategy pursues narrow criminal offences targeting specific enumerated harms. The comprehensive administrative frame is the AI Act's shape; the narrow criminal frame is the DEFIANCE Act's shape. Movements often run both threads simultaneously through different specialist coalitions.

Distinct from consumer-facing AI use disclosure mandate. Disclosure mandates require operators to identify AI to consumers; criminal-code amendments punish specific perpetrators of specific AI-enabled acts. Both are AI-specific rule-changes; disclosure is a broad civil obligation on operators, criminal amendments are narrow criminal liabilities on perpetrators.

Fed by survivor-led testimony as evidence. Non-consensual synthetic imagery and AI-generated CSAM legislation have advanced primarily where survivors of the harm have testified in legislative hearings and organised as the campaign's public face. Survivor-led narrative is what has repeatedly moved these bills through legislatures otherwise reluctant to legislate on AI at all, and criminal-code amendments in this space are more dependent on survivor-testimony organising than any adjacent AI-policy campaign.

In tension with strategic litigation against algorithmic state decisions. Strategic-litigation coalitions frequently include organisations skeptical of criminal-law expansion on prison-abolition and prosecutorial-discretion grounds; the strategies coexist in the movement but their coalitions overlap only partially, and criminal-code amendments touching enforcement-technology questions (biometric identification, predictive analytics for prosecution) can produce direct coalition conflict with the litigation strategy's opposition to algorithmic state decision-making.