AI, Identity, and Liability: What Sora 2's Backlash Means for Product Counsel
When OpenAI rolled out Sora 2, it triggered a clear legal warning shot. The tool allowed use of names, images, and likeness by default, and the response from public figures, estates, and major agencies was immediate.
The pushback worked. OpenAI shifted from opt-out to opt-in for NIL (name, image, likeness), and publicly backed federal legislation on digital replicas, including the proposed NO FAKES Act. For legal teams, the lesson is simple: design legal guardrails up front, or the market will force your hand later-loudly.
Why Sora 2 Became a Legal Flashpoint
The core issue is control of identity in a world where a few prompts can recreate a person's voice, face, and persona. That's not a niche problem-it's a systemic one for entertainment, advertising, sports, and any product with user-generated content.
OpenAI initially pitched Sora as empowering-"you are in control of your likeness end-to-end." Then the incidents hit. Actor Bryan Cranston flagged unauthorized generations. Estates pushed back on the use of figures like Dr. Martin Luther King Jr. Families asked the public to stop circulating AI images of loved ones, including the daughters of Robin Williams and George Carlin.
Industry Reaction: Consent, Credit, and Compensation
Creative Artists Agency warned Sora "poses risk to creators' rights" and questioned whether creators would be credited and paid. United Talent Agency called unconsented use of IP and likeness "exploitation, not innovation." SAG-AFTRA leadership raised the specter of mass misappropriation without strong guardrails.
OpenAI responded with "regret for these unintentional generations," first adopting opt-out, then moving to opt-in for NIL. The company also announced plans for more granular controls and signaled support for federal regulation. That's a full pivot-from permissive defaults to consent-first and pro-legislation.
The NO FAKES Act: A Federal Floor for NIL
Reintroduced in April 2025 as a bipartisan Senate effort, the NO FAKES Act targets a clean, practical aim: protect creators' and performers' control over their voice and visual likeness while keeping room for responsible innovation. It would also streamline a messy patchwork of state laws by establishing, for the first time, a federal right of publicity in voice and visual likeness.
If enacted, it could supplant or partially preempt state rules, delivering clearer obligations for product design and enforcement. For counsel, that means fewer jurisdictional surprises and stronger arguments for consent-first defaults. Track the bill's progress and text on Congress.gov.
Practical Guardrails Product Counsel Can Ship Now
Don't wait for litigation or regulation to set your standards. Build consent and context into the product from day one.
- Prompt filtering: Detect and block prompts that seek to generate or edit content with identifiable people (e.g., "make a video of [celebrity] endorsing [brand]").
- Consent (opt-in by default): No use of a person's NIL unless the person (or estate) has affirmatively granted permission.
- Context analysis: Separate informational or educational use (e.g., "who is the CEO of X?") from commercial or endorsement scenarios. Treat the latter as high risk without explicit consent.
These controls reduce misuse, curb secondary liability exposure, and give you a better story with regulators, partners, and press. They also create space for legitimate uses without chilling effect claims.
Liability Posture: Don't Rely on Hopes and Edge Cases
Some defenses might apply to user actions, but they're fact-specific and untested at scale in AI contexts. Even if a user is primarily liable, developers can face secondary liability theories. The fastest way to de-risk is to prevent the misuse in the first place.
Work with counsel early to pressure-test your defaults, permissions stack, and enforcement. Treat NIL controls like privacy and safety: fundamental, not features.
Action Items for Legal Teams
- Move default policies from opt-out to opt-in for NIL and sensitive IP.
- Implement prompt filtering and context-aware gating for endorsements, ads, and commercial use.
- Publish clear rights-holder processes for consent, removal, and appeals.
- Audit model outputs and user reports; keep logs that demonstrate good-faith prevention and response.
- Track federal efforts like the NO FAKES Act and prepare to align product terms and UX accordingly.
For Public Figures and Creatives
If you're concerned about misuse of your NIL or other IP, speak with counsel experienced in safeguarding IP portfolios and reputations. Map what's vulnerable, decide where to opt out, and consider strategic licensing where it makes sense.
AI will keep testing the line between expression and individual control. The smart move-for developers and creatives-is clear rules, consent-first design, and counsel in the loop from the start.
Your membership also unlocks:
