In what may be one of the more audacious internal experiments in Silicon Valley’s ongoing love affair with artificial intelligence, a team of Uber engineers built an AI-powered digital replica of their own chief executive, Dara Khosrowshahi. The project, which surfaced publicly in late February 2026, offers a fascinating window into how large technology companies are testing the boundaries of generative AI — not just for customers, but for their own corporate hierarchies.
The AI version of Khosrowshahi was reportedly designed to simulate the CEO’s communication style, decision-making tendencies, and even his mannerisms in digital interactions. According to TechCrunch, the project was developed internally by Uber engineers, raising immediate questions about the practical applications — and ethical implications — of creating AI doppelgängers of corporate leaders.
From Hackathon Experiment to Boardroom Conversation
The genesis of the AI Khosrowshahi appears to have been an internal hackathon or innovation sprint, the kind of event that large tech firms regularly host to encourage creative thinking among their engineering ranks. Uber has long encouraged such projects, and the company’s engineering culture has produced a number of internal tools and prototypes that eventually influenced its consumer-facing products. But an AI clone of the CEO is a different kind of output entirely — one that blurs the line between technological showcase and organizational provocation.
As reported by TechCrunch, the digital Khosrowshahi was trained on publicly available data, including interviews, earnings calls, internal memos, and other communications attributed to the CEO. The resulting model could reportedly field questions and offer responses that mimicked Khosrowshahi’s tone and reasoning. While the project was not sanctioned as an official Uber product, it nonetheless attracted significant internal attention and, eventually, external scrutiny.
The Technology Behind the Digital CEO
Building a convincing AI replica of a specific individual requires more than simply feeding a large language model a collection of speeches and emails. Engineers working on such projects typically fine-tune foundation models — often built on architectures similar to those powering OpenAI’s GPT series or Meta’s LLaMA — using carefully curated datasets that capture not just what a person says, but how they say it. Sentence structure, vocabulary preferences, rhetorical habits, and even the cadence of responses all factor into the training process.
In Uber’s case, the engineers reportedly used a combination of publicly available transcripts from Khosrowshahi’s numerous media appearances, his posts on social media, and transcripts from Uber’s quarterly earnings calls, which are publicly filed with the Securities and Exchange Commission. The result was an AI agent that could approximate the CEO’s voice with enough fidelity to be recognizable to colleagues who interact with him regularly. Whether the model captured Khosrowshahi’s actual strategic thinking — as opposed to a surface-level imitation of his communication patterns — remains an open question.
Corporate AI Clones: A Growing Trend With Uncomfortable Implications
Uber’s experiment is not occurring in a vacuum. Across the technology industry, companies have been exploring the creation of AI avatars and digital twins of real people for a variety of purposes. Startups like Synthesia and HeyGen have built businesses around generating AI video avatars for corporate communications. Microsoft has integrated AI-powered “copilots” into its enterprise software that can draft emails and summarize meetings in a user’s personal style. And several firms have experimented with AI agents that can stand in for executives during routine internal Q&A sessions or onboarding processes.
But creating an AI version of a sitting CEO introduces a distinct set of complications. For one, there is the question of authority: if an AI Khosrowshahi issues guidance or answers a strategic question, does that carry the weight of an actual directive from the CEO’s office? Even if the tool is clearly labeled as a simulation, the psychological effect on employees who interact with it could be significant. Research from Stanford University’s Human-Centered AI Institute has shown that people tend to ascribe more authority and trustworthiness to AI systems that are modeled on real authority figures, compared to generic chatbots.
Legal and Ethical Minefields
The legal dimensions of AI cloning are evolving rapidly and remain largely unsettled. In the United States, several states have enacted or proposed legislation addressing the unauthorized use of a person’s likeness through AI, often in the context of deepfakes. California, where Uber is headquartered, passed legislation in 2024 aimed at protecting individuals from nonconsensual AI-generated replicas, though much of that law was targeted at political advertising and explicit content rather than corporate use cases.
For a company like Uber, the internal nature of the project may offer some legal insulation — Khosrowshahi, as the company’s CEO, presumably has some awareness of and potential control over how his likeness is used within his own organization. But the precedent is uncomfortable nonetheless. If engineers can build an AI clone of the CEO, what stops them from building one of a middle manager, a board member, or a departing executive whose communication style the company wants to preserve? The slope from innovation to overreach is steep, and corporate governance frameworks have not yet caught up with the technology.
Khosrowshahi’s Own Stance on AI
Dara Khosrowshahi has been publicly vocal about Uber’s AI ambitions. Under his leadership, the company has invested heavily in machine learning for ride pricing, route optimization, fraud detection, and customer service automation. In recent earnings calls, Khosrowshahi has described AI as central to Uber’s strategy for improving margins and expanding into new verticals, including freight logistics and autonomous vehicle partnerships.
His reaction to being digitally cloned by his own engineers has not been extensively documented in public statements, though TechCrunch indicated that the project was received with a mix of amusement and curiosity within the company. Khosrowshahi, who has cultivated a reputation as a more approachable and less combative leader than his predecessor Travis Kalanick, may view the project as a testament to his engineers’ creativity. But even the most open-minded CEO might pause at the idea of an AI system speaking on his behalf, however unofficially.
What This Means for the Future of Executive Communication
The broader implications of Uber’s experiment extend well beyond one company’s internal hackathon. As AI models become more capable of mimicking specific individuals, the line between authentic executive communication and AI-generated approximation will become increasingly difficult to draw. Investor relations, internal corporate messaging, and even regulatory filings could eventually be touched by AI systems trained on the voices of specific leaders.
This raises fundamental questions about accountability and authenticity. When a CEO’s AI clone answers an employee’s question about company strategy, who is responsible for the accuracy of that answer? If the AI model reflects outdated thinking — trained on statements the CEO made before a strategic pivot, for instance — the potential for confusion is real. And in a world where corporate communications are subject to legal discovery, the existence of an AI system that generates CEO-attributed statements could create novel liabilities.
The Talent War Angle: Why Engineers Build What They Build
There is another dimension to this story that deserves attention: the motivations of the engineers themselves. In a fiercely competitive market for AI talent, engineers at major tech companies are constantly looking for projects that will sharpen their skills, build their portfolios, and attract attention — both internally and externally. Building an AI clone of your CEO is precisely the kind of high-profile, technically challenging project that can elevate an engineer’s standing within a company and in the broader industry.
Uber, like its peers, is locked in an ongoing battle to recruit and retain top AI researchers and engineers. Allowing — or at least not discouraging — ambitious internal projects like the AI Khosrowshahi may serve a dual purpose: it keeps talented engineers engaged while also stress-testing the company’s AI capabilities in novel ways. Whether this particular project yields any lasting product innovation remains to be seen, but its value as a recruiting tool and morale booster should not be underestimated.
Where Uber Goes From Here
Uber has not announced any plans to productize the AI Khosrowshahi or to expand the concept into other areas of its operations. The project, for now, appears to remain a one-off demonstration of what is technically possible. But the fact that it was built at all — and that it garnered enough attention to make headlines — suggests that the idea of AI-powered executive avatars is no longer the stuff of science fiction. It is an engineering problem that has, at least in prototype form, been solved.
For Uber’s competitors, investors, and the broader corporate world, the message is clear: the technology to create convincing digital replicas of real business leaders exists today, and the only barriers to its widespread adoption are organizational, legal, and ethical — not technical. How companies choose to address those barriers will say a great deal about the kind of AI-augmented corporate culture that emerges over the next several years.