A version of this article appeared in CNN Business’ Nightcap newsletter. Sign up for free to receive it in your inbox here.
new york
—
“The world is in danger,” the former head of Anthropic’s safeguard research team warned as he headed for the exit. OpenAI researchers, who are also considering a withdrawal, said the technology “has the potential to manipulate users in ways we don’t have the tools to understand, much less prevent.”
They are part of a wave of artificial intelligence researchers and executives who are not just leaving their employers, but are sounding the alarm loudly when they leave, calling attention to what they see as red flags.
Silicon Valley is known for its high turnover rate, and the recent turnover comes as market leaders like OpenAI and Anthropic compete for IPOs, which could both accelerate growth and invite increased scrutiny of their operations.
In recent days, a number of high-profile AI staffers have decided to stop developing AI, with some explicitly warning that the companies they worked for were moving too fast and downplaying the technology’s shortcomings.
Zoe Hitzig, who has been an OpenAI researcher for the past two years, announced her resignation in an essay for the New York Times on Wednesday, citing “deep reservations” about OpenAI’s new advertising strategy. Hitzig, who warned of ChatGPT’s potential to manipulate users, said the chatbot’s archive of user data is based on “medical fears, relationship issues, beliefs about God and the afterlife,” and poses ethical dilemmas because people believe they are chatting with a program that has no intentions.
Hitzig’s criticism comes after technology news site Platformer reported that OpenAI had disbanded its “Mission Alignment” team, which was created in 2024 to further the company’s goal of ensuring that all of humanity benefits from the pursuit of “artificial general intelligence” (hypothetical AI capable of human-level thinking).
OpenAI did not respond to a request for comment.
Also this week, Mrinanku Sharma, head of Anthropic’s safeguards research team, posted a cryptic letter on Tuesday announcing his decision to leave the company and warning that “the world is at risk.”
Sharma’s letter only vaguely mentioned Anthropic, the company behind the Claude chatbot. He did not give a reason for his resignation, but said, “It was clear to me that it was time to move on.” “Throughout my time here, I have seen again and again how difficult it is to let our values truly govern our actions.”
Anthropic told CNN in a statement that it appreciates Sharma’s efforts to advance AI safety research. The company noted that he was not the head of safety or responsible for broader safety measures within the company.
Meanwhile, at xAI, two co-founders quit within 24 hours this week, announcing their departure at X. That leaves just half of xAI’s founders remaining with the company, which is merging with Elon Musk’s SpaceX to create the world’s most valuable private company. At least five other xAI staff members announced their departures on social media in the past week.
It was not immediately clear why the latest xAI co-founders left, and xAI did not respond to requests for comment. Musk said in a social media post Wednesday that xAI has been “reorganized” to accelerate growth, but that “unfortunately, we have had to part ways with some people.”
While it’s not uncommon for high-level talent to bounce around in emerging industries like AI, the scale of turnover in xAI in such a short period of time is striking.
The company faced a global backlash over its Grok chatbot, which was allowed to generate non-consensual pornographic images of women and children for weeks before the team stepped in to stop it. Grok also had a tendency to generate anti-Semitic comments in response to user prompts.
Other recent departures highlight tensions between some researchers concerned about safety and management eager to generate revenue.
On Tuesday, the Wall Street Journal reported that OpenAI fired one of its top security executives after he expressed opposition to the rollout of an “adult mode” that would allow pornographic content on ChatGPT. OpenAI fired its head of safety, Ryan Beiermeister, for discriminating against male employees, but Beiermeister told the Journal the accusations were “completely false.”
OpenAI told the Journal that her firing was unrelated to “any issues she raised during her employment.”
High-level defections have been part of the AI story since ChatGPT hit the market in late 2022. Shortly thereafter, Jeffrey Hinton, known as the “Godfather of AI,” quit his job at Google and began evangelizing what he considered the existential risks posed by AI, including massive economic shifts in a world where many “no longer know what is true.”
There are many predictions of the apocalypse. This includes among AI executives who have a financial incentive to hype the capabilities of their products. One of those predictions went viral this week, with HyperWrite CEO Matt Schumer posting a nearly 5,000-word screed claiming that the latest AI models are already making some tech jobs obsolete.
“We are telling you what has already happened in our work and warning you that you are next,” he wrote.
