Artificial intelligence has already entered cinema. It can intervene in writing, previsualisation, dubbing, visual effects, animation, editing and the reconstruction of a face or voice. The question is no longer whether a film is pure or contaminated by technology: cinema has always been a technological art form. The decisive issue is when a tool continues to extend a human choice and when it begins to occupy the place of the author or performer who would receive the recognition.
The rules published by the Golden Globes for its 84th edition address this threshold without imposing a general ban. The use of AI, including generative AI, does not automatically make a work ineligible. A film may compete if creative direction, artistic judgement and authorship remain primarily human throughout production. It is a pragmatic position: technology is already part of workflows, but the machine’s industrial performance cannot be presented as the creative achievement of a person.
The distinction is therefore not between films “with AI” and films “without AI”. That division would be too simple and quickly impossible to enforce. The real line runs between assistance and substitution. A model may propose variants, accelerate a task, correct a defect or enable an image designed and controlled by an author. It becomes problematic when it determines the essential form of the work without an identifiable person taking responsibility for the decisions. The criterion is not the amount of technology, but where the creative centre remains.
For acting categories, the rules are especially concrete. A performance must derive primarily from the credited performer’s work; if expression, movement or vocal delivery is substantially generated or determined by AI, it is not eligible. Technical or cosmetic interventions such as de-ageing, ageing and visual modification remain possible if the underlying performance is still the actor’s. The issue is not only how long a face appears on screen, but who actually constructed that presence.
This protects more than a credit. Acting is not simply a set of features to reproduce: it emerges from intention, gesture, hesitation, rhythm, breathing and relationships with other performers. A system can recombine these elements into a convincing figure, but resemblance is not automatically an authorised performance. The Golden Globes therefore also prohibit unauthorised use of a performer’s digital likeness, voice replication or biometric data, whether or not the person is credited.
Consent becomes part of artistic integrity rather than merely a legal clause. A performer may authorise a correction, a translated voice or a visual transformation for a specific purpose; this does not mean handing over their identity indefinitely for every future production. Face and voice are now reproducible production assets, but they remain projections of a person. Without limits of duration, purpose and control, digital continuity can become appropriation.
The second pillar is disclosure. Every submitted work must describe any generative AI used anywhere in production, including alterations to a credited performer’s appearance or voice. The eligibility committee may request additional material, and missing information may lead to exclusion. Disclosure is not an admission of wrongdoing; it makes the process legible. For producers and authors, logs of tools, versions, consent and decisions become part of the film’s actual documentation.
Direction, writing, composition and animation must also originate primarily from credited people. AI may support or enhance these crafts, but not replace their core creative contribution. This is the policy’s hardest area: terms such as “primarily”, “substantially” and “supporting” are not automatic measurements. They require interpretation, comparisons among cases and a shared culture of film crafts. The rule does not remove ambiguity, but it forces productions to explain who decided and what each person can genuinely claim.
The Academy of Motion Picture Arts and Sciences has chosen language that is partly different and more categorical in some areas. For the 99th Oscars, eligible performances must be demonstrably performed by humans with their consent, while screenplays must be human-authored. The Academy also reserves the right to request information about AI use and human authorship. The two systems are not identical, but they converge on one principle: technical innovation cannot make invisible who created the work and who authorised the use of their identity.
As legal scholar Oreste Pollicino observes, these policies are becoming industrial, creative and reputational compasses. They are not state law, yet they can change production behaviour long before a court becomes involved. A technically lawful film may still conflict with the standards of an award, festival, distributor or platform. AI governance therefore also passes through private organisations that define access to the circuits where prestige, markets and cultural memory are formed.
One risk remains: vague rules may favour large productions able to sustain audits and legal advice, while independent authors face costly uncertainty. Common disclosure templates, simple logs, provenance credentials and criteria known before delivery will be needed. Transparency works when it helps people design a responsible process, not when it becomes a retrospective test based on shifting definitions.
The award remains human not because the machine must be hidden, but because cultural recognition needs to attribute a choice. To award something is to identify who imagined, selected, performed and answered for the result. AI may enter the film, even profoundly; what it cannot do is turn authorship into a nameless zone. The challenge is not to preserve a cinema without technology, but to build one in which every innovation still leaves consent, responsibility and intention visible.