Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Former OpenAI engineer William Saunders said he resigned in February 2024 because he had lost confidence that the company would take sufficient precautions as it developed increasingly capable AI. In a July interview, he compared the company’s direction to the Titanic rather than the Apollo program. That was his judgment about OpenAI’s priorities—not proof of a secret policy or an independently established finding that the company always put profit ahead of safety.
Who is William Saunders?
Saunders worked at OpenAI for about three years, doing alignment research and later working on its Superalignment team. He was one member of that team, not OpenAI’s overall safety chief. He resigned in February 2024 and later spoke publicly about why he had lost confidence in the company’s approach.
In a July interview with journalist Alex Kantrowitz, Saunders described asking himself whether OpenAI’s trajectory resembled the Apollo program or the Titanic. He said he did not want to end up working for the “Titanic of AI.” The comparison was a metaphor for what he saw as a race to build increasingly powerful systems without enough preparation for failure; it was not a prediction that an AI disaster was imminent. Futurism’s report on the interview summarizes his remarks, and a Big Technology excerpt includes the comparison.
What was the “upsetting truth”?
The phrase in the headline can sound like Saunders uncovered a hidden fact. The evidence supports a narrower account: he reached a conclusion about OpenAI’s incentives and direction. He argued that the company increasingly behaved like a product business focused on releasing newer offerings and competing, while he wanted it to take “all possible reasonable steps” to reduce the risks of advanced AI.
#1 Best Overall
Saunders did not argue that technology can be developed with zero risk. His concern was whether the company was doing enough to manage risks that might grow as AI systems became more capable. That distinction matters: his resignation establishes that one former employee lost confidence, but it does not by itself establish OpenAI’s motives or settle how safe any particular release was.
What Superalignment was meant to do
Alignment generally means making AI systems behave in ways consistent with human intentions and societal expectations. Superalignment addressed a harder, prospective problem: how to align and control future systems that could be more capable than the people supervising them. It did not mean that human-surpassing AI had already been built, nor did it establish when such systems might exist.
The team was associated with OpenAI co-founder and former chief scientist Ilya Sutskever and alignment leader Jan Leike. In May 2024, after both men left, OpenAI disbanded the Superalignment team. Reporting said its personnel were moved into other groups; the team’s dissolution should not be confused with a claim that all safety work at OpenAI stopped. Axios reported on the team’s disbanding.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallThe wider dispute over safety and speed
Saunders’s account arrived amid other departures and public criticism, but those events are distinct from his own reasons for leaving. Former and current employees raised concerns about safety resources, internal dissent, and pressure to move quickly. Those are allegations and accounts from particular people, not a complete independent audit of OpenAI’s decisions.
Rank #3
One separate dispute concerned testing for GPT-4o. Contemporaneous reporting cited insiders who described testing as compressed into roughly a week and “squeezed.” OpenAI disputed that it had cut corners; an internal source reportedly acknowledged a squeezed process while denying that safety procedures had been skipped. This contested reporting is not evidence Saunders himself supplied about GPT-4o testing. Futurism’s related report describes the competing accounts.
In a separate interview, Saunders also said he was given a non-disparagement agreement and told that refusing to sign could result in cancellation of vested equity. That is his account of his own employment circumstances; it should not be generalized to all former OpenAI employees. His interview transcript, including his description of alignment work and resignation, is available through Your Undivided Attention.
Rank #4
What OpenAI said it would change
OpenAI announced a board Safety and Security Committee on May 28, 2024, to oversee safety and security processes. On September 16, the company said the committee would become an independent board oversight committee and could delay a model release if safety concerns remained unresolved. These were the company’s announced governance measures, not proof that every criticism had been answered or that every safety question was settled.
The dates help keep the story in perspective: Saunders resigned in February; the Superalignment team was disbanded in May; OpenAI announced its committee on May 28; the Apollo–Titanic interview appeared in July; and the company outlined its later oversight structure in September. The resignation and interview were not the same event, and the governance announcements came after he had left. OpenAI’s announcements are available in its May committee statement and September safety update.
What Saunders’s resignation does—and does not—show
His decision is evidence of a serious loss of confidence by someone who had worked on alignment at OpenAI. Together with the Superalignment team’s dissolution and other employees’ public concerns, it shows that safety priorities and governance were contested inside and around the company.
But “OpenAI put profit over safety” remains a criticism, not a conclusion established by Saunders’s resignation alone. His Apollo-versus-Titanic analogy captures his fear that competition and product momentum could outrun precautions. It does not prove that every OpenAI release was unsafe, that the company had no safety program, or that an AI catastrophe was inevitable.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
