Safety researchers at major AI labs are increasingly resigning over concerns that artificial general intelligence (AGI) development is moving too fast and poses existential risks to humanity. Bilal Chughtai, a research engineer focused on AGI safety and alignment at Google DeepMind, departed in July 2026 and posted stark warnings on social media about the trajectory of AI development.

Why Are Top AI Safety Researchers Leaving?

Chughtai's resignation is not an isolated incident but part of a documented pattern of safety staff departing frontier AI labs. His work centered on alignment, the field focused on keeping AI-powered systems controllable and within human intent. In posts reported by Bloomberg, Chughtai expressed that he was "extremely concerned by the default trajectory" of AI development after witnessing it firsthand at Google DeepMind.

Chughtai

The departure carries particular weight because Chughtai worked directly on the safety challenges that labs claim to prioritize. His public warning that "AI has the potential to kill us all, and that we might be running out of time to avoid this outcome" reflects growing frustration among researchers who believe the industry is not moving cautiously enough.

Beyond Chughtai, Josh Engels, another former Google DeepMind AGI safety researcher, also left the company and joined METR, a separate AI safety organization. This pattern suggests that safety-focused talent is voting with their feet, seeking roles outside the major labs where they believe they can have more influence over risk mitigation.

What Are the Signs That Industry Leaders Are Taking These Concerns Seriously?

Despite the exodus of individual researchers, there are emerging signals that executives at rival AI companies recognize the need for caution. Anthropic CEO Dario Amodei published an essay titled "We Must Pace the Frontier," in which he argued that AI progress has accelerated "drastically faster" than anticipated and called for stronger safety testing, independent evaluations, and international cooperation.

What makes Amodei's position noteworthy is the public alignment it has generated from competitors. Sam Altman, CEO of OpenAI, stated, "I agree with Dario that we need to pace the frontier," while Elon Musk simply responded, "Dario is right". This convergence across fiercely competitive organizations represents a rare moment of consensus in an industry typically defined by rivalry.

Sam Altman, CEO of OpenAI

"AI has the potential to kill us all, and that we might be running out of time to avoid this outcome," warned Bilal Chughtai in his resignation statement.

Bilal Chughtai, Research Engineer at Google DeepMind

How Are Institutional Concerns About AI Risk Shifting the Industry Conversation?

The combination of researcher departures and executive statements marks a significant shift in how AI risk is being discussed. This is no longer a theoretical argument confined to academic papers or safety conferences. When safety researchers resign over existential concerns and the chief executives of major AI companies publicly endorse slowing development, the conversation has moved from speculative risk to institutional reckoning.

The stakes underlying this shift include multiple dimensions:

• Safety Testing: Calls for stronger safety testing protocols before models are deployed to ensure systems behave as intended and do not pose unforeseen risks.
• Independent Evaluations: Demands for external oversight and audits of AI systems by researchers outside the companies building them, reducing conflicts of interest.
• International Cooperation: Recognition that AI development cannot be governed by individual companies or nations alone, requiring coordinated global frameworks.

These institutional positions represent a departure from the move-fast-and-break-things mentality that has dominated tech for decades. The fact that safety researchers are leaving and executives are publicly calling for restraint suggests the industry is grappling with risks that cannot be ignored or solved through incremental improvements alone.

The resignations and public statements also highlight a tension within the AI industry. While some leaders acknowledge the need to slow down, the competitive pressures to develop more capable models remain intense. Whether these public commitments translate into meaningful changes in development practices remains an open question, but the shift in rhetoric itself signals that existential risk concerns have moved from the margins to the mainstream conversation among AI leaders.