Google held confidential negotiations with Disney, Universal, and Warner Bros. Discovery to license characters and film libraries for its generative models, but no agreements have been signed yet. In parallel, YouTube offered studios a program to detect AI-generated content featuring other people's faces, but negotiations were halted by a legal requirement that rights holders do not consider acceptable. The scale of potential deals is measured in billions of dollars, according to the Los Angeles Times.

image

What happened

A Los Angeles Times article, prepared by Wendy Lee and published on August 31, 2026, describes Google's confidential negotiations with Disney, Universal, and Warner Bros. Discovery studios to license characters and film libraries for the company's AI models. Agreements have not yet been signed. According to an estimate by a source familiar with similar deals, a license for a single character costs an average of about $40 million, so a contract for a catalog of 100 characters would cost billions. A separate line of negotiations concerned the likeness detection program: YouTube offered studios and talent agencies the detection of AI-generated content featuring other people's faces, but the sticking point was the requirement to waive the right to sue Google.

Context

Until recently, Google and Hollywood were exchanging legal blows: last year, Disney sent the company a cease-and-desist letter over training models on its content, and today Disney and Universal are suing Midjourney. Now Google is betting on money instead of lawsuits. In June 2026, Google invested $75 million in studio A24 and is collaborating with director Darren Aronofsky and his company Primordial Soup. The niche of a studio partner was freed up after the closure of the OpenAI Sora app: because of this, Disney's $1 billion deal with OpenAI fell apart, and Google is trying to take the freed-up place.

Why this matters for the industry

For the industry, the main thing here is not the technology, but the first-ever negotiating anchor for the price of studio IP for generative models — although it is incorrect to call this a market, since there are no deals, only one anonymous estimate. If the licensing model becomes established, access to premium catalogs will turn into a game with billion-dollar checks, where startups have no entry, while a bet on original content looks relatively cheaper. The condition of waiving legal rights, which derailed the face recognition program, outlines the boundary of requirements that rights holders consider unacceptable, and likeness detection itself may become a mandatory infrastructure layer for platforms. The practical conclusion for product teams today: do not build an architecture that depends on legally generated recognizable characters, and proactively build guardrails into generative tools.

Why this matters for users

Right now, nothing changes for readers: there is no legal way to create a video with a recognizable studio character through Google's generative tools, nor are there open prices or products. If at least one deal is closed, generative video with famous characters will become a legal feature of Google platforms, and rights holders will receive a new source of income and a share of YouTube advertising. Until then, reports of negotiations should be perceived as an early stage of bargaining, not as a completed fact.

What is still unknown / limitations

The reliability of the picture is limited: NBCUniversal denies active negotiations, and Google, Disney, and Warner Bros. Discovery declined to comment. The only figure for the cost of a character is an estimate by one anonymous source familiar with similar deals, not a public price list, so it is incorrect to call it a market price. Only a description of the function is known about likeness detection: there are no publications, datasets, or accuracy metrics, and it is impossible to assess the quality of the detector. Finally, it does not follow from the material whether a potential license would cover model training or only on-demand generation — this key question remains open.

Sources

Author

Look at AI, editorial team