Let me tell you a story about a company that asked a famous actress to voice their AI assistant, got told no, and then launched a voice that sounded suspiciously like her anyway. Then, when called out, their CEO tweeted "Her" with a winking reference to the movie where Scarlett Johansson voices an AI. Then they pulled the voice and issued a statement saying it was a coincidence.
If this were a case study in my brand strategy class, I would call it "How to Destroy Trust in Three Easy Steps."
WeWork's S-1 filing used the word 'community' 150 times while describing a real estate sublease business. The gap between language and reality was the entire cautionary tale.
Step one: signal that you do not respect boundaries.
OpenAI reportedly approached Johansson twice about voicing their assistant. She said no. They moved forward with something close enough to make everyone uncomfortable. Whether or not the voice was technically derived from her performance (they claim it was a different actress), the optics are devastating. The message received by the public was: this company will approximate what it cannot have permission for.
For a company building technology that raises enormous questions about consent, intellectual property, and the use of creative work without permission, this is the single worst narrative you could generate. It confirms every fear that artists, writers, and creators have about AI companies.
Step two: be cute about it.
Sam Altman's "Her" tweet was the strategic equivalent of winking at a judge during your own trial. It connected the dots between Johansson and the product in a way that either revealed the intent or demonstrated catastrophically poor judgment about how it would be perceived. Either reading is bad.
Step three: retreat to legal technicality.
"It was a different actress" may be legally defensible. But brand trust does not operate on legal standards. It operates on perception. And the perception is: they wanted Scarlett Johansson's voice, she said no, and they got as close as they could without getting sued. That perception is now permanently attached to the OpenAI brand whether it is technically true or not.
Why this matters for all of us:
We are in the early innings of AI brand building. The companies leading this space are establishing the trust foundations that will either support or undermine their products for the next decade. Every interaction with the public is either depositing or withdrawing from the trust account.
OpenAI just made a massive withdrawal. Not because of the voice itself, but because of the pattern it reveals: a company that treats consent as an obstacle to be navigated rather than a value to be honored.
For brands operating in AI, adjacent to AI, or simply existing in a world where AI ethics are becoming a consumer concern: your handling of consent, credit, and intellectual property is now a brand attribute. It is not just legal risk management. It is positioning. Consumers are watching how you treat creators, and they are drawing conclusions about how you will treat them.
The lesson is not complicated: When someone says no, the answer is no. And if you choose to get clever about it, do not be surprised when the internet sees through you immediately. Trust is built in drops and lost in buckets, and OpenAI just kicked over a very large bucket.
The lesson is not complicated: When someone says no, the answer is no.