As its valuation surges past $800 billion, OpenAI also has a row of lawsuits and regulators hanging over its head. For this company, the law is one of the biggest sources of uncertainty beneath that lofty valuation.

This piece lays out, all at once, the main legal battles OpenAI is facing: Elon Musk’s lawsuit, the New York Times copyright case, Europe’s rulings, and the regulatory scrutiny of its nonprofit conversion plus the pressure of the EU AI Act. If you want to get the full picture of the company first, you can start with What kind of company is OpenAI.

One thing to remember up front: most of these cases aren’t over yet, and the right move now is to track them rather than treat them as settled.

The Musk lawsuit: a first-instance win, but not on the point that matters

Elon Musk is one of OpenAI’s co-founders and left the board in 2018. In 2024 he sued OpenAI, Sam Altman, and Microsoft, with the core accusation being that OpenAI betrayed its founding promise to “benefit humanity through a nonprofit” and turned itself into a profit machine. At one point his demand for damages climbed into the tens of billions of dollars.

In May 2026, the jury reached its verdict: Musk’s claims were time-barred under the statute of limitations and could not proceed, giving OpenAI a first-instance win. But there’s a key detail that often gets overlooked here: the court did not rule on the substantive dispute of whether OpenAI actually breached its nonprofit mission; it simply blocked the case on timing grounds. Musk immediately said he would appeal.

In other words, OpenAI won on procedure, not on the merits. That core question remains unresolved. The backstory of this lawsuit is inseparable from OpenAI’s governance history, so it’s worth reading alongside The Sam Altman boardroom upheaval.

OpenAI’s copyright lawsuits aren’t just one case; here we pick out three main threads.

The New York Times case: The New York Times accuses OpenAI and Microsoft of using its news to train models without authorization, and alleges ChatGPT’s output can reproduce the original text nearly verbatim. OpenAI counters that training is “fair use” and that ChatGPT is not a substitute for the news. The case is currently in the discovery phase, with the court even requiring OpenAI to hand over a large volume of de-identified ChatGPT conversation records. It hasn’t settled and hasn’t been finally decided, making it one of the single biggest legal risks worth watching over the long term.

The authors’ class action: A group of authors led by the Authors Guild has also sued, arguing that ChatGPT’s output is highly similar to their books. The court has rejected OpenAI’s motion to dismiss, letting the case proceed to trial.

The European GEMA case: In late 2025, a court in Munich, Germany ruled that OpenAI’s unauthorized use of song lyrics to train models constituted infringement, the first such ruling in Europe and a significant one. OpenAI has announced an appeal, and this first-instance verdict is not yet final.

The nonprofit conversion: regulators gave a green light, but with conditions

OpenAI’s shift from a nonprofit to a for-profit structure was itself an undertaking that required regulatory approval. In 2025, the attorneys general of California and Delaware each stated, subject to conditions, that they “do not object” to its restructuring.

The conditions include: the nonprofit OpenAI Foundation must retain control over the for-profit public benefit corporation (PBC), hold a substantial proportion of equity, and have the power to appoint directors. Critics argue, however, that this arrangement still has loopholes, and that the for-profit arm could in practice still dominate decision-making. Combined with the appeal of Musk’s lawsuit, the controversy over this structure hasn’t truly come to a close.

The EU AI Act: Europe’s tightening grip

In Europe, OpenAI also has to contend with the EU AI Act. Because of the enormous scale of its training, its models are classified as general-purpose AI (GPAI) with “systemic risk,” carrying heavier obligations than ordinary models.

These obligations include: conducting model evaluations and red-teaming, reporting serious incidents within set time limits after they occur, disclosing energy efficiency, and publishing summaries of training data. The EU’s enforcement powers are expected to take effect in August 2026, at which point violations could be fined up to a set percentage of global annual revenue. For a company operating worldwide, this is a compliance line that can’t be taken anything less than seriously.

Regulatory hot spots: youth safety and capability risk

Beyond the cases above, there are two regulatory hot spots worth mentioning.

The first is consumer and youth safety. The U.S. Federal Trade Commission (FTC) has opened an investigation into several AI companies, including OpenAI, focusing on the impact of AI chatbots on minors, especially around mental health. OpenAI has responded that it will roll out parental account linking and related safety notification mechanisms. This is a challenge the entire industry faces together, and OpenAI is not an isolated case.

The second is safety tiering of model capabilities. OpenAI has a “Preparedness Framework,” and when a model’s capabilities in areas such as biology, chemistry, or cybersecurity are internally rated as high risk, additional controls kick in, such as opening access only to vetted safety researchers. Notably, these capability ratings are currently mostly OpenAI’s own self-assessments, lacking independent third-party verification, which critics see as an institutional gap.

Penchan’s take

Put these lawsuits and regulatory matters side by side and you’ll find they actually revolve around the same core: a company that has grown too fast and too big is colliding with “the existing legal and social rules.” The boundaries of copyright, the original intent of being a nonprofit, the protection of minors, cross-border regulation, every one of these lines is still being drawn.

For readers, the most practical reminder is this: the vast majority of these cases aren’t finished. A first-instance verdict, a dismissal, a denied motion to dismiss, these are all just nodes along the way, not the end. When you read legal news about OpenAI, first ask, “Is this the final ruling?” and you’ll often avoid mistaking the process for the conclusion.