Breaking News That Never Happened

Generative AI is in the news again this week. 

First, Jacob Coxon, an ex-Anthropic researcher, resigned this week because he said that his employer and other AI companies are not developing the technology responsibly.

Second, Anthropic’s CEO Dario Amodei said that AI companies needed to slow their development to give safety measures a chance to catch up. This announcement came a few days after Coxon’s resignation and in the wake of a slew of negative press and consumer pushback. 

Amodei’s suggestions for increased safety measures were ambiguous and, frankly, too little too late.

These safety measures are in response to fears over creating AI systems that we are not prepared to control. We have already created AI systems that we cannot control. 

On Friday, my coworker sent me a NewsBreak article claiming our company had launched a new program. It was dated August 10, 2026, but neither of us had heard about this program launching. Before we sounded the alarm bells, I read through the article, checked the credibility of the publication, reviewed the sources, and searched to see if there was any other coverage of this program launch.

There wasn’t. 

This statement was at the bottom of the article: This article includes content summarized and organized with AI tools, which can make mistakes. 

“Mistakes” is an understatement. 

The article pitched the program as being new. It gave readers the impression that it was something they could sign up for today. 

We did facilitate this program at one point in time. 

That point in time was October of 2020. 

NewsBreak used AI to scrape the internet for story ideas, found some press releases and documents, and generated a story using the information. A misleading and false story. 

This isn’t the first time NewsBreak has published erroneous information. Reuters reported that between 2021 and 2024, at least 40 stories generated with AI included inaccurate information that affected the local communities the organization claims to serve, including spreading fake news of a mass shooting in a small New Jersey town. NewsBreak responded by taking down the article and putting a disclaimer on its website noting that some of the information in its articles may be inaccurate. Instead of fixing the problem, NewsBreak effectively pushed the responsibility of fact checking back on the consumer—a role that is the responsibility of the publisher, not the reader. 

There are many, many layers to this situation that disturb me. I hold a bachelor’s degree in journalism and worked for a variety of monthly and weekly newspapers and magazines before, during, and after college. Providing factual information is the top priority for any journalist. 

There are times when journalists won’t get it right—of course. They are human. This is why corrections exist. News organizations publish corrections when they get something wrong, either in print or online. In modernity, this looks like a note at the beginning or the end of an online article noting the changes that were made.

NewsBreak didn’t even attempt to make corrections. 

If this doesn’t unsettle you, it should.

We are long past the need for safeguards to be put in place with artificial intelligence. Maybe AI hasn’t taken over the internet yet, but should the worst-case scenario really be the only thing we guard against? 

Generative AI and the people who pioneer it are already causing harm. Amodei’s proposed safety measures aren’t going to repair the harm that has already been done, nor will they focus on safe and ethical use of existing AI tools, only the development of new ones. 

Safeguards are about preventing future harm, while corrections are about acknowledging existing harm and mitigating its effects. It’s time for AI companies and the organizations that use these tools to make corrections and take responsibility for the harm they’ve caused.