Microsoft, Adobe, and OpenAI commit to protecting their clients from intellectual property lawsuits, but this guarantee does not apply to everyone.
Last week, at the first OpenAI developers' conference, CEO Sam Altman introduced a set of legal protections that the company calls the "Copyright Shield."
"The new policy means that we will intervene and defend our customers and pay the costs if you face copyright infringement lawsuits," he said.
When asked why OpenAI is confident in legal protection for clients, especially when copyright infringement lawsuits from writers, music labels, and comedians are hitting Silicon Valley, as technology companies use internet content to train chatbots and image technologies, he responded that the company is very confident in its approach and wants to share that confidence with developers.
In reality, OpenAI is late to the compensation game. In June, Adobe announced that it would protect customers from intellectual property lawsuits related to the use of its artificial image generator, Firefly.
In September, Microsoft followed suit with the Copilot Copyright Commitment, promising to compensate customers facing lawsuits for the use or distribution of materials created by artificial intelligence in programs like Windows, Word, PowerPoint, and its code generator, GitHub Copilot.
Last month, Google published a flexible blog post revealing legal protection for users of its artificial intelligence services. "Simply put, for you, our customers, if you are challenged based on copyright, we will take responsibility for potential legal risks."
The predictable truth, of course, is that if you read the fine print, the proposed protection is narrower than what PR suggests—and Altman's brief answer at the press conference.
Firstly, these policies apply only to commercial customers who pay for the use of services such as ChatGPT Enterprise and Firefly for Enterprise. These premium options may include an additional set of safeguards, primarily preventing the accidental use of copyrighted materials.
On the other hand, if you use a free service, such as Dall-E, and take an artwork containing images of a Mickey Mouse doppelganger, place it on a billboard, and receive a lawsuit from Disney, the protection does not apply.
The same goes for a ChatGPT user who instructs the chatbot to create a stylish new trademark containing the phrase "Just do it," and OpenAI's lawyers won't come to your aid.
Brenda Leong, a partner at Luminos Law, a firm specializing in AI issues, states that compensations apply to specific corporate models and customers of specific versions sold by these companies, which have many means of control to prevent leakage of any protected information.
It is also worth noting that the current wave of lawsuits from writers and other authors is aimed at the extensive use of training data in popular AI models, such as copyrighted books, open-source code, and web images.
Compensations provided by Adobe, Google, Microsoft, and OpenAI are intended to protect against intellectual property lawsuits related to materials obtained from AI models, and no one has sued a client or AI service user for copyright infringement caused by AI.
Nevertheless, the new policy and its accompanying pomp indeed serve their purpose: to reassure nervous companies, compelling them to embrace a new crop of enterprise-level generative AI tools. The same concerns plagued the open-source movement in its early days before companies like Red Hat deployed protective measures similar to those discussed by today's AI companies.