Explained | Why Amazon is restricting its employees from using generative AI tools like ChatGPT

In a recent email, Amazon updated its policy to restrict the use of generative AI, following the lead of Apple and Samsung. But why are these tech giants, who are reportedly working on their own generative AIs, restricting its use?

Amazon recently cautioned its employees against utilising content generated from third-party AI, such as ChatGPT, for work-related purposes. This advisory follows several internal memos in the company that iterated the same, as was reported by India Today.

This comes at a time when gaining copyright over such content has become increasingly challenging. There is also the possibility of the companies offering generative AI services asserting ownership of the generated content used in confidential documents.

This encompasses various outputs like emails, internal documents, and materials before their official launch. These could potentially be accessed, examined, and distributed by the operators of the generative AI tools.

So, one can understand why Amazon is following the same path as other tech giants like Apple and Samsung, and restricting the use of AI content. Even Microsoft, which invested heavily in OpenAI, the company that started the generative AI frenzy, temporarily revoked access to in-house generative AI tools for its employees.

Tech companies are taking precautions against AI content following several copyright cases in regard to AI generated content. An example of this is the comic book Zarya of the Dawn, which used AI-generated images from Midjourney, and thus the US copyright office revoked the image copyrights of the comic.

In a recent communication to its staff, Amazon underscored the significance of protecting sensitive information. The message stressed the need to refrain from utilising generative AI tools, particularly for tasks related to Amazon's confidential operations, despite their potential convenience. The company explicitly cautioned against disclosing any proprietary Amazon, customer, or employee data when employing third-party generative AI tools.

"While we may find ourselves using GenAl tools, especially when it seems to make life easier, we should be sure not to use it for confidential Amazon work." Amazon said, according to the publication.

These limitations are partly driven by concerns surrounding ownership rights in light of Microsoft's substantial investment in OpenAI. Microsoft's association raises the prospect that they might assert ownership over the outcomes produced by such models.

Amazon's spokesperson, Adam Montgomery, however, emphasised that the company has been working on developing its own generative AI and learning models, adding that employees use the technology on a daily basis, according to India Today. But Amazon is concerned for their confidential information and security of their documents, resulting in the imposition of such restrictions.

Amazon's warning emphasizes the company's dedication to safeguarding sensitive data and retaining authority over proprietary information amidst the rapid evolution of AI technologies.

Amazon and its AI-generated product descriptions

A recent product listing on Amazon for a dresser caught attention due to its peculiar message: “I’m sorry but I cannot fulfill this request; it goes against OpenAI use policy.”

The listing even claimed the dresser had two drawers, but the accompanying picture clearly showed three. This raises concerns about whether Amazon thoroughly reviews the products on its site and whether companies are using ChatGPT to quickly generate product descriptions without proper proofreading.

Amazon did clarify the situation stating that they were constantly improving their systems, and have removed the listing in question.

Similar issues were observed with other products, including an outdoor sectional and a bike pannier, which displayed the same OpenAI notice. Amazon has long been plagued by similar situations of fake paid reiews, fake listings, and other such incidents.

While these mislabeled products may not pose severe risks, they highlight the challenges of maintaining quality in e-commerce.