While personal challenges may arise, the realm of artificial intelligence appears to be encountering significant hurdles. The saga began with Grok’s notorious “racism glitch,” and now, SoundCloud is under fire for a questionable clause hidden within its terms of service.

The trouble can be traced back to February 2024 when SoundCloud discreetly modified its terms to include the following:

Unless a separate agreement specifies otherwise, you explicitly consent to your content being used to inform, train, develop, or serve as input for artificial intelligence or machine intelligence technologies or services as part of providing the services.

This clause went unnoticed for over a year until Ed Newton brought it to light this past week through a post on his X account. The reaction was swift and intense, with many artists and SoundCloud users protesting against the potential training of AI on their music.

However, some of the backlash seems to be misdirected; the situation is more complex than it may appear at first glance.

Understanding SoundCloud’s Intentions with AI

It’s understandable why musicians would oppose their creations being utilized to train machines that could eventually replace them. Yet, according to SoundCloud, the change in terms was not intended in that manner.

Eliah Seton, the company’s president, published an open letter on Wednesday, clarifying that the purpose of employing AI was solely for “enhancing recommendations, improving search functionality, playlisting, content tagging, and tools for fraud prevention,” specifying that the company has never used artists’ content to train AI models. This includes music production or any large language models that aim to imitate or replace original works.

Seton indicated that the misunderstanding stemmed from ambiguous language, stating, “The terms of use were too broad and lacked clarity. This led to confusion, and we take responsibility for that.”

The Ongoing Debate: Opt-in vs. Opt-out

Although SoundCloud has clarified its past uses of AI, representatives seemed noncommittal about future practices regarding user music. Marni Greenberg, Senior Vice President and Head of Communications, remarked in a statement to Verge, “If we ever consider leveraging user content to train generative AI models, we will implement clear opt-out options beforehand.”

The community’s response was clear: shouldn’t it be an ‘opt-in‘ policy?

SoundCloud eventually agreed, replying that opting in would indeed be preferable.

Revisions to SoundCloud’s Terms of Service

In his open letter, Seton detailed forthcoming adjustments to the service’s terms. The contentious AI clause is set to be rephrased to read:

We will not utilize your content to train generative AI models aimed at replicating or synthesizing your voice, music, or likeness without your explicit consent, which must be provided through an opt-in mechanism.

This means AI will not be used for duplicating or synthesizing users’ music without their consent. However, it appears SoundCloud will continue to leverage AI for recommendations, tagging, and playlisting—functions that are generally more accepted and benign.

This incident serves two valuable lessons: Firstly, companies must strive for clear communication regarding AI’s applications with users. Secondly, it is crucial to read terms of service diligently.