Artificial intelligence has developed at lightning speed in recent years—especially generative AI systems like ChatGPT, which are transforming how we get information, write, and process data. But the smarter these systems get, the more one big question stands out: Where exactly did they learn all this?

Put simply, AI doesn’t “create from nothing.” Its intelligence comes from massive amounts of material scraped from the internet—articles, books, news reports, even music and artwork. This means many AI models may have infringed on creators’ copyrights during training.

Microsoft Sued—Authors Aren’t Staying Silent

Recently, a copyright lawsuit involving Microsoft put this issue back in the spotlight. Several well-known American authors—including Kai Bird, Jia Tolentino, and Daniel Okrent—have sued Microsoft in federal court in New York.

They claim Microsoft used nearly 200,000 pirated books without permission to train its AI model, Megatron. These books were all protected by copyright, yet they became AI’s “training material.” What’s more worrying is that Megatron doesn’t just parrot the text—it can generate writing whose style, themes, and structure closely mimic the originals.

Microsoft isn’t alone in facing these allegations. Meta, Anthropic, and OpenAI have all been caught up in this wave of lawsuits, suggesting the problem isn’t isolated but rather an industry-wide challenge.

What Does the Law Say? Rulings Are Still Uncertain

Shortly before this Microsoft lawsuit, a federal judge in California ruled in a similar case involving Anthropic. That decision said using copyrighted materials to train AI might, in some circumstances, count as “fair use.” But the court also made clear that using pirated content could still be illegal.

In other words, there’s no settled answer yet on whether training AI with copyrighted material is allowed. Different states and judges may interpret the law differently. A truly consistent standard is still being worked out.

Not Just Microsoft: More Media and Authors Are Speaking Up

The Microsoft case is far from the only one. Lawsuits over AI and copyright are erupting everywhere:

- In 2023, Game of Thrones author George R. R. Martin and 16 other writers sued OpenAI, accusing it of “systematic theft” of their works.

- Later that year, The New York Times also sued Microsoft and OpenAI, claiming they misused the paper’s reporting.

- In 2024, Forbes and Wired filed their own claims against AI startup Perplexity for content plagiarism.

- And in the music industry, the Recording Industry Association of America sued two AI music companies, Suno and Udio, in June 2024 for using copyrighted music to train AI composition tools.

You can see a pattern: from authors and journalists to musicians, just about every group that relies on original creative work is growing uneasy about how AI is trained.

Some Block Access, Others Choose to Partner

Faced with this threat, media outlets are taking different approaches. Some are choosing to “block” AI: outlets like CNN, The New York Times, and Reuters have added code to their websites to stop AI models from scraping their content.

Others are choosing to “channel” it: Germany’s Axel Springer, the Associated Press, and other organizations have struck deals with OpenAI to let it use their content under agreed terms.

This underscores a key point: the progress of AI is inevitable. As a result, many organizations are shifting from blocking it to finding ways to collaborate.

For AI Developers: Risks and Responsibilities

So what does all this mean for AI developers?

First, you need to stay legally aware. Copyright law hasn’t fully caught up to the AI era yet, but a wave of new rules is coming. If you don’t follow these developments, you might end up with a model that launches only to trigger massive lawsuits.

Second, don’t focus solely on the technical side while ignoring use cases and potential risks. For example, if your AI can imitate a specific author’s style so closely it’s practically identical, you could be in serious trouble. Before training, you need to check your data sources carefully, build good filtering systems, and manage outputs responsibly.

Third, educating your users is critical. If someone uses your AI-generated text and submits it as their own original work to a publisher, you might still get dragged into liability. Clear disclaimers and user “guardrails” aren’t just for show—they’re essential.

It’s Not a Technology Problem—It’s an Ecosystem Problem

AI’s development itself isn’t the problem. The real issue is that we haven’t yet built a full “ecosystem” to support it—laws, ethics, regulations, and usage norms.

Going forward, a few things will be necessary:

1. Clarifying in law exactly what data AI can use for training—what counts as “fair use,” and what’s infringement.

2. Defining developer responsibilities clearly so that technologists don’t become scapegoats.

3. Ensuring regulators keep pace with the technology instead of always scrambling to catch up.

4. Investing in user education so that people don’t blindly treat AI-generated content as genuine or original.

In short: The AI wave is already here. But remember: when the wave passes, the ones still standing will be those who prepared their safe harbors in advance.