Rights, ethics, and disclosure

Before you create for real, know the rules — they protect you legally and ethically. This area moved from "totally unsettled" to "partially clarified but still contested" in 2026, so here's the durable picture; verify specifics for your jurisdiction.

Copyright: only human authorship is protected. US courts (Thaler, affirmed on appeal; the Supreme Court declined to review in 2026) and the US Copyright Office are clear: a work generated autonomously by AI can't be copyrighted, and prompting alone generally isn't enough to claim authorship, because you don't control how the model expresses the idea. What can be protected: your own human-authored contributions, your creative selection and arrangement of AI material, and your creative edits — but protection covers only the human parts. Practical takeaway: you can copyright your script, your edits, and how you arrange AI outputs; you generally cannot copyright a raw prompt-to-image result. Disclose AI use if you register anything.

Commercial use and indemnification. Tools differ enormously. Some (Adobe Firefly, and certain Microsoft/Google offerings) provide IP indemnification and train only on licensed data — the safe choice for business. Others assign you ownership of outputs but offer no indemnity, and open models put the legal risk on you. Free tiers often restrict commercial use entirely. Read the terms of service for anything you'll use commercially — the load-bearing details are there, not in the marketing.

Training-data lawsuits are ongoing. Suits against image, music (Suno/Udio), and text AI companies are unresolved, with some settlements and split rulings. You generally won't be liable for a tool's training data — but you can create infringing output (e.g., generating a famous copyrighted character), which is on you.

Deepfakes and consent. Don't create realistic synthetic media of real people without consent — it's an ethical line and increasingly a legal one (some deepfake laws are now enforceable). Voice cloning especially requires the person's consent.

Disclosure is now standard, sometimes required. Provenance standards (C2PA "Content Credentials," Google's SynthID watermark) are converging, platforms (YouTube, TikTok, Meta) require you to label realistic AI content, and the EU AI Act's transparency rules (from August 2026) require marking synthetic media for EU audiences. Disclose AI-generated content — it's becoming both an expectation and a rule.

The through-line: create ethically, use commercially-cleared tools for commercial work, never fake real people without consent, and disclose. Do that and you create with confidence instead of legal anxiety.