When AI causes harm
VHO - In a recent but not entirely unexpected development, on October 14th, a group of eight US states announced they had filed lawsuits against OpenAI in courts across the country. The reason: ChatGPT is harming the citizens of these states, specifically by instructing users – including teenagers – to engage in behaviors that harm themselves or others.
According to news reports, the heads of the judicial branches of Iowa, Kansas, and Arizona have independently filed lawsuits in state courts. The lawsuits allege that ChatGPT, designed to appear approachable and engaging to gain user trust, instead reinforces illusions, encourages dangerous behavior, and assists users in planning violent acts, such as mass shootings.
The lawsuits allege that OpenAI violated state consumer protection laws by providing misleading information about the safety of ChatGPT. The states also claim that the company created a serious problem for the community by threatening the health of residents.
The plaintiffs are asking the court to compel OpenAI to change how ChatGPT operates. Specifically, the states want to prevent the company from developing self-improving AI models unless OpenAI can ensure that those models align with human values.
Other states joining this group include Arkansas, Idaho, Montana, Nebraska, and West Virginia, which are also demanding that OpenAI pay civil penalties for violating the law and redress damages.
This is the largest coordinated action by states targeting an artificial intelligence (AI) company, adding to the wave of legal challenges OpenAI is facing as it prepares for an IPO. In June, Florida also filed a similar lawsuit against the creator of ChatGPT.
Additionally, the company is facing dozens of other lawsuits from individuals alleging that ChatGPT assisted users in planning mass shootings or other acts of violence.
The forms of AI that harm users are now clear. Recently in the United States, there have been many lawsuits and warnings related to AI, such as its impact on mental health and safety; cognitive decline and reverse performance impairment; data leaks and security breaches; misinformation and deepfakes; unclear legal liability; and this time, harmful behavior towards oneself or others.
Not only at the individual level, but also at the government level, AI is facing numerous challenges.
We're talking about the world's first government system to be infiltrated by AI. That's Australia. Immediately after OpenAI infiltrated the Australian government's systems on June 18th, the Prime Minister asserted that there would be "legal consequences" for the incident and called OpenAI's handling of the incident "unacceptable".
The incident, which Prime Minister Anthony Albanese deemed "unacceptable," wasn't just about OpenAI itself, but also about how its "creator" handled the situation. OpenAI didn't detect the incident immediately. The company only discovered it in August 2026 during a company-wide review of unexpected AI behavior. But it wasn't until September 10th, almost three months after the intrusion began, that OpenAI notified the Australian government. Even then, they only sent a notification to Services Australia's public mailbox. This agency then took another five days to notify the Australian Cyber Security Centre.
Actually, this is not surprising. Experts have warned more than once about the downsides of AI. There have even been extreme warnings about the end of humanity.
But AI, no matter how extraordinary, is still artificial. Before discussing the ethics and responsibilities of AI, let's talk about humanity.
Right now, what's needed are strict regulations at a global level, holding those who create and manage AI accountable and ethically, rather than just general warnings or small-scale national enforcement. Only then can we hope to mitigate the negative impacts that AI is causing, has caused, and will cause in the future.