Microsoft to shield paid-up Copilot customers from any AI copyright brawls it starts

Tough luck, freeloaders: You're on your own

Microsoft vowed on Thursday it would step in and defend paying customers if they face any copyright lawsuits for using Copilot.

Magnanimous, one might say, though another way of looking at it is this: Microsoft is offering experimental products – AI tools that generate text, code, and more - that customers are so worried will get them sued, the software giant is offering to shoulder that risk so buyers feel more confident about signing up and deploying the tech.

"To address customer concern, Microsoft is announcing our new Copilot Copyright Commitment," the mega-corp's president Brad Smith and chief lawyer Hossein Nowbar announced.

"As customers ask whether they can use Microsoft's Copilot services and the output they generate without worrying about copyright claims, we are providing a straightforward answer: yes, you can, and if you are challenged on copyright grounds, we will assume responsibility for the potential legal risks involved." 

Microsoft: China stole secret key that unlocked US govt email from crash debug dump


Microsoft promised to shield customers, and pay the costs of damages or settlements from such lawsuits, but only if plaintiffs "used the guardrails and content filters we have built into our products," and only if they are using the paid versions of the company's tools. Those that are only using the free version of Bing or GitHub Copilot will not be protected, for instance.

The Copilot Copyright Commitment covers the Microsoft 365 Copilot, which deploys generative AI tools for Word, Excel, PowerPoint, as well as its Bing Chat Enterprise and GitHub Copilot, "specifically for paid versions." We've previously covered Redmond's efforts to inject its Copilot-branded generative AI and software assistants into all corners of its realm:

"We are charging our commercial customers for our Copilots, and if their use creates legal issues, we should make this our problem rather than our customers' problem," Smith and Nowbar explained.

The boom in AI tools capable of automatically outputting content, such as text, code, images, and audio, from an input description has ignited a debate over copyright and ownership. 

Organizations building this kind of software trained their generative bots on data scraped from the internet and other sources. Books, artwork, articles, journals, code projects, and songs from authors, engineers, writers, scientists, and artists are swept up into massive datasets ingested by neural networks, allowing the resulting models to mimic human styles convincingly.

There is some alarm among creatives and those in other industries that this use of harvested information, to train highly valuable, publicly available models, falls foul of copyright law as some kind of unauthorized use. There is also a concern or feeling that the models' output is also subject to copyright in some way as it at times closely resembles individuals' work due to its inclusion in training material.

Some professionals have sued the likes of OpenAI, Google, Midjourney, and Stability AI for allegedly lifting their work without consent for training and generation, and profiting off their IP. The tech companies, however, claim that their use of copyrighted content falls under fair use, which protects transformative uses that repurposes that content. 

The legal gray zone has apparently made some people nervous about using Microsoft's generative AI tools, who are worried that they could be sued for copyright infringement.

Whether AI systems violate copyright laws at the training or output stages isn't quite clear, the US Copyright Office has meanwhile at least clarified that AI-generated content cannot be protected by copyright. Officials rejected a request to register a digital image titled "Théâtre D'opéra Spatial" created using Midjourney, for example, this week.

The image made headlines and stirred controversy last year after it won first place in an art competition. Its creator Jason Allen said he produced the image by refining an input prompt "at least 624 times" to get the desired output, which he then touched up and rescaled using software from Adobe Photoshop and Gigapixel AI.

The US Copyright Office asked Allen to refile his copyright claim excluding the parts that used Midjourney, but he declined. Since the law only recognizes works from human authorship, the office denied his application to register the image.

"After considering the application, the deposit, and Mr Allen's correspondence, the board concludes that the work contains an amount of AI-generated material that is more than de minimis and thus must be disclaimed. Specifically, the board concludes that the Midjourney Image, which remains in substantial form in the final work, is not the product of human authorship," it said. ®

PS: Microsoft's service agreement for its online products and such stuff has been updated and is coming into effect at the end of this month. The liability section of the new fine-print is interesting reading in light of the above.

More about


Send us news

Other stories you might like