Public choices / October 2026

AI copyright

A song, photo, lesson or news story can carry years of work. Creators need a practical way to know how it is used; learners and small developers need rules they can actually follow.

Open the room Twinkle seeds Twinkle 23 · video

Twinkle / policy room

Connected Twinkles

Twinkle 23 / Brainstorming scenario

AI copyright: permission, payment and a usable remedy

Make training and reuse permissions visible, give creators collective bargaining options and provide a practical route to contest misuse. Keep training, generated outputs and impersonation as separate questions.

AI training, generated outputs and cloning a person’s identity are different uses. Personal permission profiles and source-linked legal tools offer ways to make those choices and disputes understandable.

Explore the connected work

Relevance: Australian law - The source-led method for following legal threads and understanding how laws touch an ordinary life. Source repository
Australian Legal Engine - Offline code that parses, indexes and retrieves quoted statutory provisions with source citations. It does not decide a legal case. Source repository
UNGA81: Joyful Responsible Abundance - The connected proposal for capability, ownership, livelihoods, care and worthwhile futures, with a reference library. Source repository

How it could work

Require a readable training-use record and machine-readable permissions for participating publicly supported services.
Develop a creator-negotiated collective licence pilot with transparent distribution, permitted uses and withdrawal conditions.
Offer accessible complaint handling, provenance checks and review by accountable humans.

Everyday example

A photographer can see whether a service has permission to train on their work and challenge an unlicensed use without hiring a litigation team first.

Who could act and how it could be funded

Copyright reform needs Commonwealth action; states and institutions can require licensing transparency in procurement. Fund the pilot's negotiation and dispute process; licence payments and administration must be separately disclosed.

Fair-go safeguards

No blanket claim that every training use is unlawful or that a licence solves every output dispute. Protect legitimate expression, public-interest research and creators' ability to negotiate.

What would show progress

Permission records complete; creator payments traceable; dispute time and cost; corrected misuse; licence uptake.

Twinkle 23 · video Open the full room Rabbit-hole policies

Full room: ideas, tools and sources

Delivery roles: Australia's three levels of government · Victoria's parliamentary responsibilities.

Three different questions

Training permission, rights in AI-assisted outputs and affordable enforcement are separate problems. IP Australia’s 2026 report identifies licensing, certainty about generated material and less costly enforcement as current policy priorities. A paid software subscription does not settle every right in the material fed to it.

IP Australia: copyright policy priorities

Choose rules that people can use

Opt-in and collective licencesCreators can negotiate terms and payment, including through collective arrangements. Administration and bargaining power matter: a system affordable only to large firms can shut out smaller developers and independent creators.

Targeted research exceptionsA tightly defined research pathway could lower barriers to public-interest inquiry. Its scope, security, commercial reuse and effect on creator income need explicit limits; this is an option to debate, not a statement that a new exception exists.

Transparency and accessible remediesDataset provenance, rights contacts and a usable complaint process can make rights enforceable. Detailed disclosure must also handle privacy, security and legitimate confidential information. A cheap remedy still needs fair review and an appeal path.

Keep the creator in the workflow

Ready S.E.T. cultural node keeps viewing, storage, transcription, model work and publication as separate permission decisions. For a local film or song project, record who supplied the material, the agreed use, credits, model services, release version and correction contact. Local equipment alone does not grant copyright permission.

Ready S.E.T. permission model Media projects and their status Community compute and custody

Put a concrete example on the record

An artist, teacher, publisher or small developer can compare one actual workflow under each option: permission cost, payment, turnaround, privacy and remedy if something goes wrong. Send a redacted example to the relevant consultation or your federal representative; keep unpublished work and personal details out of a public submission.

PM&C’s September consultation covers large data centres and conditions for AI training, including Australian creators. Its published closing date is 9 October 2026. Use the official page to check whether submissions are still accepted; after that date, follow published outcomes and later consultation rounds.

Consultation paper and submission route

Twinkle 23: AI copyright

A public workbench

Source check: 3 October 2026. The choices here are proposals for discussion, not an adopted party platform. Follow the linked source for its jurisdiction, date and latest status.

The studio invoice meets the training bill.

An artist can sell a finished work without agreeing to every future use. The question is who can trace a training use, negotiate terms and afford a remedy.

A question for the workbench

“If my work has value to the model, let’s talk terms before it disappears into the machine.”