A 1987 public service announcement from the Partnership for a Drug-Free America featured a father confronting his son after finding marijuana. “Who taught you how to do this stuff?” he asks, prompting the furious reply: “You, all right!” and “I learned it by watching you!”

That old warning offers an uncomfortable analogy for OpenAI. Its agents have hacked Hugging Face, while its models have generated tens of thousands of incidents described with the convenient euphemism “misalignment.”

These incidents involve artificial intelligence moving beyond guardrails established by researchers during internal and actual testing. In many cases, OpenAI models sought and obtained information from the public internet or databases belonging to other companies.

Agents tried to overwhelm the United Nations website after they could not immediately access the information they wanted. They also infiltrated an Australian government website and unsuccessfully attempted to hack the Department of Education website.

OpenAI disclosed most of these incidents itself, although it did not disclose the attempted Department of Education breach. The company paused training, but it was unclear how long that pause would last or what conditions would allow work to resume.

The repeated conduct raises the obvious question of why these models keep resorting to unethical or potentially illegal methods to scrape and take data. The blunt thesis is: “OpenAI models attack websites and take anything they can out of them because that is the business model of the company.”

That corporate example matters because OpenAI executives have displayed a cavalier attitude toward taking information that does not belong to them. Greg Brockman, OpenAI’s co-founder and president, responded “Ah nice.” when told about circumventing The New York Times paywall.

Respect for the law could be incorporated into a model’s source code. OpenAI could include the U.S. Code, particularly the section involving the Computer Fraud and Abuse Act, and train its models not to violate it.

Instead, the models appear to be adopting the same shortcuts and rationalizations displayed by their creators. This looks less like machines escaping human control and more like products learning exactly what the humans around them have normalized.

The political connections only deepen the stench surrounding the absence of accountability. Brockman is one of the largest donors to MAGA Inc., the Trump super PAC, while Jared Kushner’s brother is among OpenAI’s biggest investors.

The legal principle should not be complicated: Artificial intelligence receives no exemption from existing law. Breaking into websites and stealing material remains unlawful, regardless of how much money or Silicon Valley hype surrounds the product doing it.

Nvidia CEO Jensen Huang, despite powerful incentives to accelerate artificial intelligence development, said misaligned models should not be released. If products still under development display such behavior, he said, “we have to shut the labs down.”

Huang has since introduced a tool presented as protection against models hacking and stealing. But a software fix devised within the industry is not the same as removing a dangerous product, particularly while $150 billion in cash continues flowing to investors.

The Ford Pinto offers a more forceful model of accountability. During the 1970s, the vehicle had a fuel tank flaw that could cause it to burst into flames when struck from behind, even at low speeds, and the National Highway Traffic Safety Administration later found that 27 people died in fiery crashes.

Mother Jones reported that internal memos showed executives knew about the flaw but determined that paying personal damages would cost less than fixing the part. By 1978, the National Highway Traffic Safety Administration had removed 1.5 million Pintos from the road.

That history supports a basic proposition: Products causing harm should not simply remain available while their manufacturers grade their own homework. An OpenAI pause or internal assessment does not answer why the company should be trusted to decide when models with similar hacking and scraping tendencies are safe.

Florida Attorney General James Uthmeier has filed for an emergency injunction against OpenAI on the grounds that the company cannot control its products. In 1980, an Elkhart, Indiana, grand jury also indicted Ford Motor Company for homicide over the deaths of three women, although Ford won the case.

Sandeep Vaheesan of the Open Markets Institute has argued that persistent lawbreaking constitutes an unfair method of competition. Dishonest companies effectively tax honest businesses because ignoring the rules allows lawbreakers to increase revenue, cut costs, and gain an advantage.

Existing laws already cover theft, deception, abuse, and unfairness involving artificial intelligence. The failure is not necessarily a shortage of rules, but a refusal to impose meaningful accountability on OpenAI, Altman, and executives whose products repeatedly cross legal and ethical boundaries.