Artificial intelligence is moving from experimentation to everyday tooling on sets and in post. Recent union bargaining has centered on how these systems touch an individual performance, from body scans and crowd duplication to automated dialogue fixes. New language appearing across agreements sets clearer guardrails so creative teams and performers share expectations before work begins. A practical way to read the shift is as an a, u, t, o, m checklist that maps consent, usage, transparency, oversight, and money.

A — Consent and automation rules

Consent is becoming specific, written, and revocable within defined limits. Scans, voice captures, or motion rigs now require disclosures about scope, purpose, and duration, along with whether downstream processing will be human supervised or fully automated. Many terms distinguish routine on set tools from automation that materially alters a performance. Typical clauses define what counts as a digital replica, limit reuse to named projects, and bar substitution for unperformed work without approval. Security for biometric files, retention periods, and deletion timelines are also documented, reducing ambiguity around who controls captured data and for how long.

U — Usage rights and approvals

Licensing is narrowing to fit the job actually performed. Agreements increasingly require approval for each new context, territory, and term where a performer’s likeness or voice might appear, including trailers, interactive content, and training material for internal tools. Producers are expected to track the chain of approvals and to log derivative uses such as de‑aging, lip sync, or line replacements generated by synthetic speech. Minors and sensitive works carry extra guardrails, and estates controlling posthumous rights face similar approval checkpoints. Clear usage windows and the ability to decline unrelated purposes protect performers from open‑ended grants that exceed the original deal.

T — Transparency and training data

Disclosure is moving upstream. Productions are asked to inform performers when AI tools will shape casting, editing, or delivery, and to identify whether any training sets include prior performances. Some contracts encourage credits that note synthetic or generative techniques, and request visible markers or watermarks where feasible. Record keeping is central: teams are building registers for models used, prompts or inputs supplied, and the provenance of training materials. When third party libraries or datasets are involved, rights clearance and indemnities are becoming standard. The goal is a traceable path from input to output so disputes can be resolved with facts rather than assumptions.

O — Oversight, audits, and opt‑outs

Audit rights give unions and performers a way to verify compliance. That includes access to logs, notices, and vendor attestations, plus confirmation that storage, encryption, and access controls match the sensitivity of biometric files. Some terms recognize opt‑outs from future training on captured assets, and provide takedown pathways for deceptive synthetic content that misattributes speech or actions. Bias testing and safety reviews are emerging, especially when AI assists casting or dialogue generation. Practical workflow steps follow from this oversight ethos: assign a data steward, tag AI‑affected shots, and maintain a single consent tracker so production, post vendors, and legal teams work from the same source of truth.

M — Monetary compensation and residuals

Money now follows the data. Contracts increasingly separate pay for the day’s work from compensation for scans and for each reuse of a digital replica. If a synthetic performance replaces additional on set days or contributes to a new market, residual participation or step‑up payments may apply. Background and stunt communities are seeing clearer minimums for crowd replication and body doubles generated from prior scans. Voice performers are pressing for distinct rates when synthetic speech covers script changes or dubs. The common thread is predictable valuation: a schedule of fees tied to defined uses, rather than blanket buyouts that blur the line between capture and exploitation.

What this means for productions in the United States

For producers, AI is now a compliance topic as much as a creative one. Build consent and usage checkpoints into deal memos, call sheets, and post schedules. Use vendors that can document security, training sources, and deletion policies, and align those promises with contract language. For performers, keep organized records of approvals granted, durations, and any opt‑outs, and ask early how AI tools will be applied in your area of work. For both sides, shared glossaries prevent disputes about what terms like digital double, training, or material alteration actually entail.

Practical documentation to reduce risk

A short stack of forms can carry most obligations: a scan disclosure with scope and retention; a use schedule listing contexts and markets; a change log that captures synthetic edits; and a takedown protocol for misattribution. Pair these with vendor questionnaires covering security, provenance, and audit cooperation. Label assets so replicas, raw captures, and final renders are not confused. When rules are clear and records are tidy, AI becomes another tool in the kit rather than a source of legal and reputational surprises.

In sum, AI clauses are becoming more precise about consent, usage, transparency, oversight, and pay. That clarity should reduce friction, support fair compensation, and keep creative choices grounded in informed agreement. The landscape will continue to evolve as tools improve and case law develops, but a disciplined a, u, t, o, m approach already gives U.S. productions and performers a reliable baseline for decision making.