Inside OpenAI, the world’s most valuable startup, a question has metastasized from whisper to open wound: Can Sam Altman actually be trusted to lead a company building what he himself calls the most consequential technology in human history?
The answer, according to a growing chorus of current and former employees, board members, and close observers, is no.
A sweeping investigation by Ars Technica published in April 2026 crystallizes what has been building for years: a pattern of behavior by Altman that has eroded confidence among the very people who know OpenAI best. The reporting draws on interviews with insiders who describe a CEO whose public persona — earnest, measured, concerned about AI safety — bears diminishing resemblance to the executive they encounter behind closed doors. The portrait that emerges isn’t of a villain, exactly. It’s of something potentially more dangerous in the context of artificial intelligence development: a leader whose relationship with the truth is situational, whose promises shift with strategic convenience, and whose consolidation of power has outpaced every governance mechanism designed to check it.
This is a company that began as a nonprofit with a mission to ensure artificial general intelligence benefits all of humanity. It is now valued at north of $300 billion. That transformation didn’t happen by accident. And the tension between OpenAI’s founding ideals and its commercial imperatives runs directly through one person.
The trust deficit has concrete origins. Former board members have described to multiple outlets a pattern in which Altman provided incomplete or misleading information during critical decision-making moments. This was, in fact, the central issue behind his dramatic firing by the board in November 2023 — an event that lasted all of five days before employee revolt and investor pressure forced his reinstatement. The board members who voted to remove him were themselves removed instead. The message was clear: Altman was untouchable.
But the 2023 episode didn’t resolve the underlying concerns. It buried them.
According to the Ars Technica reporting, insiders describe an environment where Altman’s statements to employees frequently contradict his statements to the board, to investors, or to the public. One former senior employee characterized conversations with Altman as exercises in real-time narrative construction — not lying in the traditional sense, but a fluid reshaping of facts to suit the audience and the moment. “He doesn’t think he’s being dishonest,” this person said. “That’s what makes it so effective and so dangerous.”
The safety question looms largest. OpenAI’s original charter committed the organization to prioritizing safety over commercial interests. Altman has repeatedly affirmed this commitment in congressional testimony, media interviews, and company all-hands meetings. Yet the company’s actual trajectory tells a different story. The nonprofit board that once governed OpenAI has been sidelined in favor of a for-profit structure. Safety researchers have departed in waves — most notably Ilya Sutskever, the co-founder and chief scientist who was among those who voted to fire Altman, and Jan Leike, who led the superalignment team before resigning in May 2024 with a public statement that safety had become “secondary to shiny products.”
Leike’s departure was not an isolated incident. It was a signal.
Since then, the exits have continued. Multiple members of OpenAI’s safety and alignment teams have left, several joining competitor Anthropic, which was itself founded by former OpenAI researchers Dario and Daniela Amodei who departed over precisely these governance concerns. The brain drain is not random. It follows a pattern: people closest to the safety mission leave, people aligned with rapid commercialization stay or are brought in.
Altman’s defenders — and there are many, particularly among investors and the Silicon Valley establishment — argue that this framing is unfair. Building safe AI, they contend, requires massive resources. Massive resources require commercial success. Commercial success requires a CEO willing to make hard tradeoffs. In this telling, Altman isn’t abandoning safety; he’s funding it. Microsoft’s multibillion-dollar investment, the partnerships with enterprise customers, the consumer subscription revenue from ChatGPT — all of it feeds back into research, including safety research.
There’s some truth to that argument. But it conveniently elides the governance question. Who decides which tradeoffs get made? Under the original structure, a nonprofit board with no financial stake made those calls. Under the current structure, Altman does. And the people who were supposed to provide oversight — the independent directors, the safety-focused researchers, the mission-driven co-founders — are largely gone.
The financial dimensions are staggering. OpenAI’s recent funding rounds have valued the company at levels that rival the largest public technology firms. Altman himself initially claimed he had no equity in OpenAI, a statement he made repeatedly and that contributed to his image as a selfless steward of a world-changing mission. That claim has since been complicated. Reporting from The Wall Street Journal revealed that Altman was in discussions to receive a substantial equity stake as part of the company’s restructuring — a stake that could be worth billions. The gap between the public narrative of sacrifice and the private reality of potential enrichment is precisely the kind of discrepancy that fuels insider distrust.
So what does this mean for the industry?
OpenAI is not just another startup. Its models power applications used by hundreds of millions of people. Its technology is being integrated into Microsoft’s products, into healthcare systems, into financial services, into government operations. The question of whether its CEO can be trusted isn’t academic. It has direct implications for how AI development is governed during a period when the technology’s capabilities are advancing faster than any regulatory framework can keep pace with.
The broader AI industry is watching closely. Anthropic has explicitly positioned itself as the safety-first alternative, with a governance structure designed to prevent the kind of power consolidation that has occurred at OpenAI. Google DeepMind operates within the constraints — and resources — of Alphabet’s corporate governance. Meta has taken the open-source route, distributing its models publicly and arguing that transparency is itself a safety mechanism. Each of these approaches has flaws. But none of them concentrate as much unchecked authority in a single individual as OpenAI’s current structure does in Altman.
Congressional interest has intensified. Senators from both parties have raised questions about OpenAI’s governance transition, and the attorney general of Delaware — where OpenAI’s nonprofit is incorporated — has been examining whether the conversion to a for-profit entity complies with state law governing charitable assets. California’s attorney general has also weighed in. These aren’t abstract legal proceedings. They go to the heart of whether billions of dollars in assets originally dedicated to a public mission can be redirected to enrich private shareholders, including potentially Altman himself.
The cultural dynamics inside OpenAI compound the problem. Current employees describe a company where internal dissent has become increasingly risky. OpenAI’s controversial offboarding agreements — which at one point included provisions that departing employees could lose vested equity if they spoke critically about the company — generated significant backlash when they became public in 2024. Altman said he was not aware of the specific provisions and that they would not be enforced. But the fact that such clauses existed at all speaks to an organizational culture where loyalty to leadership is prized and scrutiny is discouraged.
Former employees have described a dynamic in which Altman is exceptionally skilled at making each person he speaks with feel like a trusted insider — while simultaneously telling different people different things. This isn’t a minor interpersonal quirk. In a company making decisions about technology that could reshape economies, labor markets, and the balance of geopolitical power, the integrity of internal communication is a governance issue of the highest order.
And then there’s the question of AGI itself. Altman has been more willing than almost any other industry leader to discuss the possibility that artificial general intelligence — systems that match or exceed human cognitive abilities across all domains — could arrive soon. He has framed this as a reason for urgency, for scale, for the kind of rapid capability development that OpenAI has pursued. But critics argue that this framing serves a dual purpose: it justifies the breakneck pace of commercialization while simultaneously positioning Altman as the indispensable leader for an existential moment. If AGI is coming, the argument goes, you need someone bold enough to build it and wise enough to control it. Altman has cast himself in both roles.
The problem is that the people who’ve worked most closely with him on safety aren’t convinced he’s suited for either.
Recent reporting from The New York Times has documented how OpenAI’s board, reconstituted after the 2023 crisis, has been largely deferential to Altman. The new directors include accomplished figures from business and technology, but the board lacks the adversarial independence that characterized — however briefly — its predecessor. The people who fired Altman believed they were fulfilling their fiduciary duty to the nonprofit mission. They were punished for it. The lesson was not lost on their replacements.
This creates a feedback loop. Altman gains more control. Governance weakens. The people who might push back leave or are sidelined. Altman gains more control. And with each cycle, the gap between OpenAI’s stated mission and its operational reality widens.
None of this means OpenAI’s technology isn’t impressive. It is. ChatGPT remains the most widely used AI assistant in the world. The company’s latest models demonstrate capabilities that were science fiction five years ago. OpenAI’s engineering talent, despite the departures, remains formidable. The question isn’t whether OpenAI can build powerful AI. It’s whether the company’s governance is adequate to the responsibility that comes with doing so.
Altman’s supporters often invoke a familiar Silicon Valley archetype: the visionary founder whose intensity and occasional rough edges are the price of genius. Steve Jobs was difficult. Elon Musk is erratic. Jeff Bezos was demanding to the point of cruelty. The implication is that building transformative companies requires leaders who operate outside conventional norms of corporate behavior.
But there’s a flaw in the analogy. Apple made phones. Amazon sold books and then everything else. Tesla made cars. These are products. They can be recalled, regulated, competed against. The technology OpenAI is building — by its own leadership’s admission — could be qualitatively different. If you believe, as Altman says he does, that AGI poses existential risks alongside its benefits, then the governance standards for the organization building it should be higher than those applied to a consumer electronics company. Not lower.
The irony is sharp. Altman himself has made this argument. He has testified before Congress about the need for AI regulation. He has called for international governance frameworks. He has spoken eloquently about the risks of concentrated power in AI development. And yet within his own organization, power has never been more concentrated, oversight has never been weaker, and the people who raised alarms have never been more marginalized.
“The problem is Sam Altman” — those words, used by an insider quoted in the Ars Technica investigation, carry a weight that extends beyond one company. They point to a structural failure in how the AI industry governs itself during a period of extraordinary technical acceleration. The mechanisms that were supposed to ensure accountability — nonprofit governance, independent boards, safety-focused research teams, public commitments — have each been weakened or circumvented. Not by external forces. By the person they were designed to constrain.
What happens next is uncertain. The legal challenges to OpenAI’s restructuring could impose external accountability that internal mechanisms have failed to provide. Congressional action, while slow, could establish regulatory frameworks with real teeth. Competitive pressure from Anthropic, Google, and others could force OpenAI to demonstrate — not just assert — its commitment to safety. Or the status quo could hold. Altman could continue consolidating authority, the company could continue growing, and the trust deficit could continue widening until some future crisis forces a reckoning.
The 2023 board firing was supposed to be that reckoning. It wasn’t. It was a dress rehearsal that ended with the wrong lesson: not that governance matters, but that it can be overridden. The next test may not come with a five-day window for reversal. And the stakes, as the technology grows more powerful, will only be higher.
For now, the people inside OpenAI who worry about their CEO’s trustworthiness are left with an uncomfortable reality. They work for a company whose mission they believe in, building technology they find genuinely exciting, led by a person they don’t fully trust. That contradiction is unsustainable. The only question is how — and when — it resolves.
WebProNews is an iEntry Publication