Why documenting creative process matters now

by Diego Lopes

In 2024 Cannes started asking on submission forms whether the work used artificial intelligence. In 2025 the US Copyright Office published detailed guidance on how AI-assisted works should be declared. In 2026 the EU AI Act came into force, classifying AI-generated content as a category requiring disclosure.

The question stopped being "should I use AI?" — and became "how do I prove what I did and what the AI did?".

The problem isn't AI. It's the lack of proof.

Picture three real scenarios:

  1. You submit a screenplay to a festival and get the question: "Which parts were AI-generated?". You remember using AI to review subtext in scene 12 and to generate dialogue in act 2. But you rewrote both. How do you prove it?

  2. You sign a contract with a production company. A new clause requires declaring AI use. You remember using it in some places — but three months later, you can't reconstruct it. How do you protect your payment?

  3. A co-writer challenges your contribution. In legal dispute, a lawyer asks whether you "actually wrote" or "just refined AI output". How do you answer?

In all cases, the answer requires process record, not generic declaration.

Telos doesn't declare authorship. It records what happened.

The architectural decision of Telos was explicit: the system draws no conclusion about authorship. It doesn't say "this script is 67% human". It issues no seal. It calculates no authorial merit score.

What it does: records every operation — who generated, who refined, how much time passed between versions, which AI suggestion was accepted or rejected, which scene was born from prompt versus deep human revision.

When you need to export:

  • PDF/A "Creative Process Record" — conservative, factual language to paste in festival form or share with lawyer.
  • RFC 3161 timestamp (optional, FreeTSA) — cryptographic proof that document existed on that date.
  • lineage.json — raw data for technical experts.

The decision of how to declare stays with you. Telos just provides the material.

What this changes day-to-day

Before Telos, "going back and reconstructing the process" was impossible. You'd open a Google Doc, write, delete, generate in ChatGPT, copy, modify. In the end, the final text remained — without trail.

With Telos, each scene has its own Git-like history. You can:

  • See how many AI cycles passed through each scene
  • Compare your first version of scene 7 with the final
  • Demonstrate that you rewrote 80% of what AI generated
  • Recover old versions you discarded (and regretted)

This changes the risk calculation. You can use AI with more confidence because you know the record is being made by default.

We're not the first, but we're the first serious one

Other systems added "history" as afterthought — versions saved every N minutes without distinguishing human from AI. This doesn't answer the regulatory question.

Telos was built from the start with lineage as first-class citizen. Each operation (scene created, refined, AI-generated) becomes a BranchOp with explicit category (AI generation, human refinement of AI, human composition, AI rejection). When you export, this data becomes a readable table.

It's not "spreadsheet feature". It's the laboratory notebook the screenwriter needs to navigate the next decade.

Next steps

If you don't use Telos yet: join the waitlist. For existing users: the /provenance page explains in detail what we record and what we don't.

For those still considering: read the Manifesto first. It's short, it's honest. If the thesis doesn't convince you, better not to use it.