For years, Sam Altman has cultivated a public persona defined by benevolent futurism. As the CEO of OpenAI, he presents himself not merely as a tech executive, but as a steward of humanity’s most powerful invention. He speaks frequently of universal basic income, the democratization of intelligence, and the existential necessity of safety guardrails. However, cognitive scientist and AI critic Gary Marcus argues that this carefully curated image is collapsing under the weight of accumulated evidence. In his biting analysis titled “Breaking: Sam Altman’s Greed and Dishonesty,” Marcus dismantles the narrative of Altman as an altruistic leader, suggesting instead that a pattern of manipulation, broken promises, and aggressive power consolidation has come to define the OpenAI chief’s tenure.
The central tension lies between OpenAI’s founding mission and its current operational reality. Established in 2015 as a non-profit research lab, the organization pledged to build artificial general intelligence (AGI) that benefits all of humanity, unconstrained by the need to generate financial returns. Yet, the trajectory has shifted dramatically toward profit maximization and product dominance. Marcus points to the restructuring of the company into a “capped profit” entity—a move that allowed for multi-billion dollar investments from Microsoft—as the original sin that began eroding the company’s ethical foundation. This structural pivot created conflicting incentives, pitting the original safety-focused charter against the demands of investors seeking massive returns.
The Boardroom Coup and Transparency Failures
Doubts regarding Altman’s candor reached a fever pitch in November 2023, when the OpenAI non-profit board abruptly fired him. The board’s official statement was vague but damning, asserting that Altman was “not consistently candid in his communications.” While Altman was quickly reinstated following employee pressure and investor maneuvering, the specific reasons for the board’s loss of confidence were never fully aired publicly. Marcus emphasizes that this was not a standard corporate power struggle but a desperate attempt by the governance body to exercise its duty. The subsequent removal of safety-conscious board members, including Helen Toner and Tasha McCauley, signaled a victory for commercial interests over the oversight mechanisms originally designed to keep the CEO in check.
Former board member Helen Toner later revealed specific instances of manipulation, alleging that Altman had withheld information regarding the release of ChatGPT and had provided misleading narratives about the company’s safety processes. Marcus highlights these events to illustrate a recurring theme: when oversight conflicts with Altman’s agenda, the oversight is removed. The reconstitution of the board with industry heavyweights, such as Larry Summers, effectively defanged the non-profit’s ability to restrain the for-profit arm. This consolidation of power removed the internal friction that was supposed to serve as a safety valve for AGI development.
The Exodus of Safety Researchers
Internal dissent has not been limited to the boardroom. A significant number of high-profile safety researchers have departed the company, citing fundamental disagreements with leadership’s priorities. Among the most notable exits were Ilya Sutskever, the company’s chief scientist and co-founder, and Jan Leike, a key figure in the “Superalignment” team. Leike’s resignation was accompanied by a public statement declaring that safety culture and processes had taken a backseat to “shiny products.” Marcus seizes on this exodus as proof that the individuals best equipped to understand the risks of AI no longer trust Altman to manage them responsibly.
The dissolution of the Superalignment team—which had been promised 20% of the company’s compute resources to solve the problem of controlling superintelligent AI—further contradicts Altman’s public assurances. Reports indicate that the team never received the promised resources, leaving them unable to fulfill their mandate. For Marcus, this broken promise is emblematic of a wider strategy: talk loudly about safety to appease regulators and the public, while quietly starving safety initiatives of the resources they need to function. The dismantling of the very team dedicated to preventing catastrophic outcomes suggests that speed and market share take precedence over caution.
Equity Clawbacks and Legal Intimidation
Perhaps the most damaging revelation regarding Altman’s treatment of employees involves the controversy over off-boarding agreements. Reports surfaced that OpenAI required departing employees to sign rigorous non-disparagement agreements (NDAs) that never expired. Failure to sign these documents, or any violation of their terms, could result in the company clawing back the employee’s vested equity. In the world of Silicon Valley startups, vested equity represents the primary form of compensation for early employees, often amounting to millions of dollars. Threatening to seize this property effectively gagged former staff, preventing them from criticizing the company’s safety practices or leadership decisions.
When this practice was exposed, Altman claimed ignorance, stating he was unaware the provision existed and that it had never been enforced. Marcus treats this defense with extreme skepticism. As CEO, Altman’s signature appeared on the very documents in question. The idea that a chief executive would be unaware of the standard terms governing employee separation—particularly terms that grant the company such immense leverage over former staff—strains credulity. Marcus argues that this incident reveals a ruthlessness incompatible with the image of a benevolent leader. It suggests a culture of fear designed to silence dissent and bury potential whistleblowers under the threat of financial ruin.
The Voice Appropriation Controversy
The pattern of disregarding consent extended beyond internal employees to the public sphere, most visibly in the dispute with Scarlett Johansson. OpenAI released a voice for its ChatGPT audio mode, named “Sky,” which sounded uncannily like the actress. Johansson released a statement clarifying that Altman had approached her months prior, asking to license her voice for the system. She declined the offer. Despite her refusal, OpenAI proceeded to release a voice that was virtually indistinguishable from hers. To make the intent clear, Altman tweeted a single word on the day of the release: “her,” a reference to the Spike Jonze film in which Johansson voices an AI assistant.
Marcus cites this incident as a glaring example of entitlement and dishonesty. When confronted with the backlash, OpenAI paused the use of the voice and claimed it was never intended to sound like Johansson, attributing the similarity to a different voice actor. This denial contradicted Altman’s own tweet and his documented attempt to hire her. For Marcus, this episode encapsulates the “ask for forgiveness, not permission” ethos that governs the company. It demonstrates a willingness to appropriate likenesses and intellectual property, only retreating when faced with significant legal or public relations consequences.
Regulatory Capture and Market Dominance
Beyond individual controversies, Marcus analyzes Altman’s broader political maneuvering. The OpenAI CEO has spent considerable time in Washington, D.C., testifying before Congress and meeting with world leaders. While he ostensibly calls for regulation to ensure AI safety, Marcus argues that this is a classic strategy of regulatory capture. By advocating for complex, expensive compliance burdens, large incumbents like OpenAI can effectively pull up the ladder, making it impossible for open-source projects or smaller startups to compete. The goal is not merely safety, but the construction of a regulatory moat that secures OpenAI’s monopoly.
This tactic aligns with what Marcus perceives as Altman’s greed. While Altman famously claimed to have no equity in OpenAI, subsequent investigations revealed he holds significant stakes in related ventures and investment funds that benefit directly from OpenAI’s success. Furthermore, recent discussions about restructuring OpenAI to allow Altman to take a direct equity stake undermine the narrative of his financial disinterest. The push for massive capital injections, including the pursuit of trillions of dollars for chip manufacturing infrastructure, points to an ambition for total market dominance rather than a cautious, collaborative approach to AGI development.
The Erosion of Public Trust
The accumulation of these events paints a portrait of a leader who operates with a distinct lack of integrity. Marcus contends that trust is the most critical currency in the development of artificial intelligence. If the public cannot trust the leading AI developer to be honest about equity agreements, respectful of creative rights, or transparent with its own board, how can they trust that same leader with the development of systems that may reshape the global economy and human existence? The discrepancy between the high-minded rhetoric of saving the world and the sordid reality of legal threats and deception creates a credibility gap that is becoming impossible to ignore.
Gary Marcus’s critique serves as a warning that the governance of powerful technology cannot rely on the presumed virtue of a single individual. The dismantling of safety teams, the silencing of critics, and the manipulation of public perception suggest that the safeguards meant to protect society from rogue AI are being systematically removed to clear the path for commercial expansion. As OpenAI continues to race toward more advanced models, the character of the man at the helm becomes a central risk factor. The evidence, Marcus suggests, indicates that Altman’s primary allegiance is to his own ambition, leaving the public to bear the risks of his unbridled acceleration.
WebProNews is an iEntry Publication