The maverick who once pushed Elon Musk to the breaking point in a face-to-face confrontation has also left OpenAI.

In May 2026, at the federal court in Oakland, California, the trial scene of the lawsuit filed by Elon Musk against Sam Altman.

Bradley Wilson, a lawyer for OpenAI, walked up to the judge and handed over a small object. It was a small golden sculpture with a white stone base, crafted in a highly realistic style.

At first glance, it might be mistaken for the FIFA World Cup Trophy, but upon closer inspection, it turned out to be the rear half of a donkey: two legs, a rump, and a tail.

An inscription was carved on the base:

Joshua Achiam, never stop being a jackass for safety.

The lawyer attempted to present this object as formal evidence to the jury. The judge's reaction, recorded by multiple media outlets, consisted of just one sentence: "I don't want it."

In the end, the golden donkey rump was not admitted as evidence by the court. The judge stated that it could be considered for contextual reference but did not wish to accept it as an official exhibit. Later, OpenAI's counsel did not show the physical object to the nine jurors.

Nevertheless, it fulfilled its purpose on the internet, making the name Joshua Achiam known worldwide. Along with the absurd story between him and Elon Musk.

Let's rewind eight years back. In February 2018, in a conference room at OpenAI, around 50 to 60 employees were seated. Elon Musk was delivering his farewell speech, as he had decided to step down from the board due to conflicts over AI talent competition between Tesla and OpenAI.

Joshua Achiam, then 25 years old and only a few months into his job, raised his hand in the meeting room and interrupted Musk's departure speech.

He sounded like he was racing toward AGI. His plan to pursue AGI at Tesla was clearly unsafe and reckless.

This was his verbatim account in the 2026 courtroom. And Musk's response at that time was also recorded word for word by him:

He snapped and called me a jackass.

Eight years later, seated on the witness stand, when Musk's lawyer tried to make excuses for him, suggesting that Musk might have only used that term to push him "out of his comfort zone".

Achiam directly rejected that narrative, stating in court:

I don't think that was why he called me that. I think he was just upset that he had been challenged.

In short, one can picture the scene: a young man in his twenties, who had only joined the company as an intern in 2017, at an all-staff meeting, in front of dozens of colleagues, confronted the world's richest man, and was called a jackass in return.

This anecdote was later made into a physical trophy by his colleagues. The two people who gave him the golden donkey rump statue were Dario Amodei, the founder of the well-known Anthropic.

The other was David Luan, who later went to Google, founded his own company ADEPT, then led Amazon's AGI lab, and announced his departure in February this year to pursue a new AI project.

Dario Amodei (left) and David Luan (right)

Eight years on, the people who presented him that trophy have all become leaders of their own ventures. And he himself has just left OpenAI.

From Ensuring AI Safety to Envisioning the AI Future

On September 25, 2024, Sam Altman announced the establishment of the Mission Alignment Team, appointing Joshua Achiam as its head.

In the post announcing the appointment, the departure of three senior OpenAI executives shocked the industry: CTO Mira Murati, Chief Research Officer Bob McGrew, and Vice President of Research Barret Zoph.

The Mission Alignment Team, with a grand-sounding name, was actually a small team of 6-7 people. Its task was clear: to ensure that OpenAI's internal operations align with its stated mission of "ensuring that AGI benefits all of humanity".

Yet the team lasted less than two years, just 16 months.

In February 2026, without any public announcement, OpenAI quietly disbanded the Mission Alignment Team. The seven members of the team were reassigned to various departments across the company with no clear public information about their new roles. When questioned, an OpenAI spokesperson made a very classic statement:

The Mission Alignment project was a support function to help employees and the public understand our mission and the impact of AI. That work continues throughout the organization.

In other words: the team tasked with reminding everyone of the mission is gone, and that responsibility has been spread across the entire organization.

As for Joshua Achiam himself, his "new role" became a position that arguably never truly existed in the history of human enterprises: Chief Futurist.

The irony of this title is that he used to be responsible for keeping the company safe, and now he was tasked with imagining what the future might hold.

According to reports from Wired, as OpenAI prepares for its public listing, the list of departing employees continues to grow.

Jan Leike, who once co-led OpenAI's Superalignment team and was also committed to researching how to keep advanced AI models under human control, joined Anthropic in 2024.

Jan Leike was named one of The 100 Most Influential People in AI 2024

Miles Brundage, who led policy research, and Steven Adler, who led research on dangerous capabilities of AI models, both left OpenAI to establish non-profit organizations advocating for strict safety standards at AI labs.

Andrea Vallone, who led OpenAI's research on how ChatGPT should respond to users' mental or emotional distress, also joined Jan Leike's team at Anthropic at the end of 2025.

Five months after the disbandment of the alignment team, in July 2026, Achiam informed his colleagues: he was leaving.

The Probability of "Human Extinction Within 10 Years" Is One in a Million

If you thought Joshua Achiam was a tragic hero, a safety idealist let down by the company, his X account might surprise you.

He has been even more scathing toward the most extreme voices in the AI safety community than Musk's "jackass" remark.

On March 29, 2023, Achiam launched a series of attacks on Eliezer Yudkowsky, a prominent figure in the AI safety field, on X:

Eliezer is going to get AI researchers killed one day. His calls for extreme violence have no place in the AI safety community.

We are long past the point where we should take him seriously, even as a charming fanfiction author.

His essays are completely disconnected from all real efforts in the field, and almost all his specific technical predictions over the past two decades have been wrong.

Earlier, he had given an extremely low estimate for the probability of near-term human extinction: a 1 in 1 million chance that everyone would die within 10 years.

He said that anyone who claims the probability of "AGI intentionally killing everyone in the next 10 years" is 5-10% or higher is "definitely not thinking clearly".

And in recent controversies, he has maintained his consistent maverick stance: he supports AI safety and regulation, but resents that the safety faction always positions itself on a moral high ground.

Meanwhile, during the May 2026 court hearing, he testified that he had sold at least $10 million worth of OpenAI shares and still held shares valued at "tens of millions of dollars".

After being called a jackass to his face by Musk, he earned at least ten million dollars from the company's stock.

More interestingly, his personal homepage, a web page about a self-proclaimed geek, lists his hobbies as follows:

• Reading and writing science fiction/fantasy novels
• Having a knowledge of "Stargate SG-1" that "exceeds normal reasonable limits"
• "Pacific Rim is the greatest movie of all time."
• Life goals: "Ensuring that the long-term future of humanity is good" and "Achieving faster-than-light space travel"

The World Is in on the Secret Now

In his public resignation letter, Joshua Achiam wrote:

The world is in on the secret now and it feels possible to work on the mission from outside the walls of a frontier lab.

These words were very dignified, but by the time he left, a large number of his former colleagues had already departed — Jan Leike went to Anthropic, Miles Brundage and Steven Adler each founded non-profit organizations, and Andrea Vallone also joined Anthropic.

In just two years, OpenAI has disbanded or scaled down at least three teams related to safety and preparedness: Superalignment, AGI Readiness, and Mission Alignment.

What is the "secret" behind "The world is in on the secret"?

Perhaps it is that OpenAI is no longer a lab fighting for AGI safety, but a tech company preparing for an IPO. Perhaps it is that he finally realized that maintaining safety principles within a company is becoming increasingly difficult.

Or perhaps it is what he has always believed: that AI will eventually reach the AGI/superintelligence stage, and that AI will reshape society, not just the software industry.

He does not know exactly what he will do next. But he said he will continue "making this vision real".

This article is from WeChat Official Account "APPSO", authored by APPSO (who discovers tomorrow's products), and published with authorization from 36Kr.