Sam Altman Further Distances Himself From OpenAI Following Ex-Employee Safety Concerns
Sam Altman has stepped down from OpenAI's Safety and Security Committee Photo by Nishimura / Getty Images.
Key Takeaways
- OpenAI has appointed a new independent Safety and Security Committee.
- Previously, CEO Sam Altman sat on the committee, prompting concerns about potential conflicts of interest.
- The reshuffle follows the departure of safety-focused employees who were frustrated with the company’s direction.
As of August 2024, nearly half of OpenAI’s artificial general intelligence safety team had reportedly left the firm. Comments from company insiders suggest the issue of safety has driven a wedge between different camps.
In the latest sign that OpenAI remains divided on the matter, CEO Sam Altman has stepped down from the company’s Safety and Security Committee. Going forward, the committee will become an “ independent Board oversight committee,” OpenAI said in a blog post on Monday, Sept. 16.
Newly Independent Safety Committee
The decision to update the structure of the security committee follows a 90-day internal review of safety and security-related processes and safeguards. Establishing independent governance of the committee was listed as the first of five recommendations.
The new oversight body will be chaired by Zico Kolter, Director of the Machine Learning Department at Carnegie Mellon University. Other members will include Quora co-founder and CEO Adam D’Angelo, U.S. Army General and former head of the National Security Agency (NSA) Paul Nakasone, and Nicole Seligman, a former executive vice president of Sony.
OpenAI said the new committee will oversee the safety and security processes guiding model development and deployment, “including having the authority to delay a release until safety concerns are addressed.”
Other safety recommendations being acted upon include enhancing cybersecurity measures to address evolving threats, promoting transparency by sharing safety work through system cards, collaborating with external organizations for independent safety testing, and unifying safety frameworks across teams to ensure consistent oversight during model development.
Ex-Employees Voice Security Concerns
The overhaul of OpenAI’s oversight process follows the news that as many as half of the firm’s safety researchers had left the company. According to one such ex-employee, Daniel Koktajlo, the wave of departures was driven by researchers “individually giving” after their work was “increasingly marginalized.”
OpenAI staff in the parallel field of AI alignment have also voiced their dissatisfaction with the company sidelining what they view as critical work.
When OpenAI’s head of Superalignment Jan Leike left in May, he blamed disagreements with the company’s leadership over what OpenAI’s core priorities should be. At the same time, Leike’s fellow superalignment proponent and OpenAI co-founder Ilya Sutskever also left to start a new venture, Safe Superintelligence.
In public comments made by former staff, some criticism of Sam Altman’s leadership is often implied. This could explain why the company is increasingly showing independence from its CEO.
Sam Altman Has a History of Being Fired
Altman’s first major withdrawal from a leadership role at OpenAI occurred last year, when he was temporarily ousted by board members including Sutskever.
Since then, Altman has stepped down as head of the company’s venture capital arm and has now recused himself from a role on the safety committee. This pattern highlights a recurring theme in his career, throughout which he has been fired on multiple occasions.
Shortly after Altman was reinstated as the CEO of OpenAI in November, the Washington Post reported that he had previously been ousted as President of Y Combinator, an allegation that both he and the startup accelerator’s founder, Paul Graham, have refused to comment on.
According to the Post, the official story – that Altman left Y Combinator to focus on OpenAI – masks a major fallout that happened behind the scenes.
OpenAI CEO Accused of Self-Interested Leadership
Citing company insiders, the report claims Graham, who appointed Altman as President and acted as his mentor, fired his protege over concerns that he was putting his own interests ahead of Y Combinator’s.
Similar charges of self-interest would later surface in the wake of last year’s OpenAI controversy, yet Altman insists that his tenure as CEO is subject to the same checks and balances as any other corporate leadership role.
“I continue to not want super-voting control over OpenAI. I never have,” he recently stated.
Mixing Personal and Professional Relationships
Alongside concerns over his loyalty to other investors, Altman’s time at Y Combinator was marked by his tendency to personally invest in startups the firm incubated.
Since his departure, Y Combinator has cracked down on the dubious practice of restricting employees from making their own separate deals with promising startups.
However, Altman appears to have continued mixing personal and professional relationships at OpenAI.
OpenAI Startup Fund Raises Eyebrows
While he hasn’t invested any money himself in the OpenAI Startup Fund, Altman was listed as its legal owner until April this year.
This bizarre ownership arrangement raises a number of questions. Most importantly, why does OpenAI or the OpenAI Foundation not own the venture capital vehicle that bears their name?
In a statement, an OpenAI spokesperson explained that the ownership structure was only ever meant to be a temporary arrangement: “We wanted to get started quickly and the easiest way to do that due to our structure was to put it in Sam’s name,” they said .
While legal ownership of the fund has now been transferred to partner Ian Hathaway, Altman’s unusual role is part of a pattern of irregularities that characterize the increasingly complex structure of the now-sprawling AI empire he oversees.
Ilya Sutskever Left OpenAI to Raise $1B for AI Safety Venture SSI in Less Than 3 Months
OpenAI Faces Crisis as Nearly Half of AI Safety Team Walks Out, Insider Claims
